Authors: Jiaqing Zhang, Miguel Contreras, Jessica Sena, Andrea Davidson, Yuanfang Ren, Ziyuan Guan, Tezcan Ozrazgat-Baslanti, Tyler J. Loftus, Subhash Nerella, Azra Bihorac, Parisa Rashidi
Abstract: Patient mobility monitoring in intensive care is critical for ensuring timely interventions and improving clinical outcomes. While accelerometry-based sensor data are widely adopted in training artificial intelligence models to estimate patient mobility, existing approaches face two key limitations highlighted in clinical practice: (1) modeling the long-term accelerometer data is challenging due to the high dimensionality, variability, and noise, and (2) the absence of efficient and robust methods for long-term mobility assessment. To overcome these challenges, we introduce MELON, a novel multimodal framework designed to predict 12-hour mobility status in the critical care setting. MELON leverages the power of a dual-branch network architecture, combining the strengths of spectrogram-based visual representations and sequential accelerometer statistical features. MELON effectively captures global and fine-grained mobility patterns by integrating a pre-trained image encoder for rich frequency-domain feature extraction and a Mixture-of-Experts encoder for sequence modeling. We trained and evaluated the MELON model on the multimodal dataset of 126 patients recruited from nine Intensive Care Units at the University of Florida Health Shands Hospital main campus in Gainesville, Florida. Experiments showed that MELON outperforms conventional approaches for 12-hour mobility status estimation with an overall area under the receiver operating characteristic curve (AUROC) of 0.82 (95\%, confidence interval 0.78-0.86). Notably, our experiments also revealed that accelerometer data collected from the wrist provides robust predictive performance compared with data from the ankle, suggesting a single-sensor solution that can reduce patient burden and lower deployment costs...
Authors: Bhargav Acharya, William Saakyan, Barbara Hammer, Hanna Drimalla
Abstract: Heart rate is a physiological signal that provides information about an individual's health and affective state. Remote photoplethysmography (rPPG) allows the estimation of this signal from video recordings of a person's face. Classical rPPG methods make use of signal processing techniques, while recent rPPG methods utilize deep learning networks. Methods are typically evaluated on datasets collected in well-lit environments with participants at resting heart rates. However, little investigation has been done on how well these methods adapt to variations in illumination and heart rate. In this work, we systematically evaluate representative state-of-the-art methods for remote heart rate estimation. Specifically, we evaluate four classical methods and four deep learning-based rPPG estimation methods in terms of their generalization ability to changing scenarios, including low lighting conditions and elevated heart rates. For a thorough evaluation of existing approaches, we collected a novel dataset called CHILL, which systematically varies heart rate and lighting conditions. The dataset consists of recordings from 45 participants in four different scenarios. The video data was collected under two different lighting conditions (high and low) and normal and elevated heart rates. In addition, we selected two public datasets to conduct within- and cross-dataset evaluations of the rPPG methods. Our experimental results indicate that classical methods are not significantly impacted by low-light conditions. Meanwhile, some deep learning methods were found to be more robust to changes in lighting conditions but encountered challenges in estimating high heart rates. The cross-dataset evaluation revealed that the selected deep learning methods underperformed when influencing factors such as elevated heart rates and low lighting conditions were not present in the training set.
Authors: Shunyu Liu, Wenkai Fang, Zetian Hu, Junjie Zhang, Yang Zhou, Kongcheng Zhang, Rongcheng Tu, Ting-En Lin, Fei Huang, Mingli Song, Yongbin Li, Dacheng Tao
Abstract: Large Language Models (LLMs) have demonstrated unprecedented generative capabilities, yet their alignment with human values remains critical for ensuring helpful and harmless deployments. While Reinforcement Learning from Human Feedback (RLHF) has emerged as a powerful paradigm for aligning LLMs with human preferences, its reliance on complex reward modeling introduces inherent trade-offs in computational efficiency and training stability. In this context, Direct Preference Optimization (DPO) has recently gained prominence as a streamlined alternative that directly optimizes LLMs using human preferences, thereby circumventing the need for explicit reward modeling. Owing to its theoretical elegance and computational efficiency, DPO has rapidly attracted substantial research efforts exploring its various implementations and applications. However, this field currently lacks systematic organization and comparative analysis. In this survey, we conduct a comprehensive overview of DPO and introduce a novel taxonomy, categorizing previous works into four key dimensions: data strategy, learning framework, constraint mechanism, and model property. We further present a rigorous empirical analysis of DPO variants across standardized benchmarks. Additionally, we discuss real-world applications, open challenges, and future directions for DPO. This work delivers both a conceptual framework for understanding DPO and practical guidance for practitioners, aiming to advance robust and generalizable alignment paradigms. All collected resources are available and will be continuously updated at https://github.com/liushunyu/awesome-direct-preference-optimization.
URLs: https://github.com/liushunyu/awesome-direct-preference-optimization.
Authors: Andrzej Dulny, Farzad Jabbarigargari, Andreas Hotho, Laura Maria Schreiber, Maxim Terekhov, Anna Krause
Abstract: We propose a 3D U-Net model to predict the spatial distribution of electromagnetic fields inside a radio-frequency (RF) coil with a subject present, using the phase, amplitude, and position of the coils, along with the density, permittivity, and conductivity of the surrounding medium as inputs. To improve accuracy, we introduce a physics-augmented variant, U-Net Phys, which incorporates Gauss's law of magnetism into the loss function using finite differences. We train our models on electromagnetic field simulations from CST Studio Suite for an eight-channel dipole array RF coil at 7T MRI. Experimental results show that U-Net Phys significantly outperforms the standard U-Net, particularly in predicting fields within the subject, demonstrating the advantage of integrating physical constraints into deep learning-based field prediction.
Authors: Fabian Galis, Darian Onchis
Abstract: In the context of unsupervised learning, effective clustering plays a vital role in revealing patterns and insights from unlabeled data. However, the success of clustering algorithms often depends on the relevance and contribution of features, which can differ between various datasets. This paper explores feature weighting for clustering and presents new weighting strategies, including methods based on SHAP (SHapley Additive exPlanations), a technique commonly used for providing explainability in various supervised machine learning tasks. By taking advantage of SHAP values in a way other than just to gain explainability, we use them to weight features and ultimately improve the clustering process itself in unsupervised scenarios. Our empirical evaluations across five benchmark datasets and clustering methods demonstrate that feature weighting based on SHAP can enhance unsupervised clustering quality, achieving up to a 22.69\% improvement over other weighting methods (from 0.586 to 0.719 in terms of the Adjusted Rand Index). Additionally, these situations where the weighted data boosts the results are highlighted and thoroughly explored, offering insight for practical applications.
Authors: Jessica Hullman, Yifan Wu, Dawei Xie, Ziyang Guo, Andrew Gelman
Abstract: Methods to quantify uncertainty in predictions from arbitrary models are in demand in high-stakes domains like medicine and finance. Conformal prediction has emerged as a popular method for producing a set of predictions with specified average coverage, in place of a single prediction and confidence value. However, the value of conformal prediction sets to assist human decisions remains elusive due to the murky relationship between coverage guarantees and decision makers' goals and strategies. How should we think about conformal prediction sets as a form of decision support? Under what conditions do we expect the support they provide to be superior versus inferior to that of alternative presentations of predictive uncertainty? We outline a decision theoretic framework for evaluating predictive uncertainty as informative signals, then contrast what can be said within this framework about idealized use of calibrated probabilities versus conformal prediction sets. Informed by prior empirical results and theories of human decisions under uncertainty, we formalize a set of possible strategies by which a decision maker might use a prediction set. We identify ways in which conformal prediction sets and posthoc predictive uncertainty quantification more broadly are in tension with common goals and needs in human-AI decision making. We give recommendations for future research in predictive uncertainty quantification to support human decision makers.
Authors: Yanxia Zhang, Francine Chen, Shabnam Hakimi, Totte Harinen, Alex Filipowicz, Yan-Ying Chen, Rumen Iliev, Nikos Arechiga, Kalani Murakami, Kent Lyons, Charlene Wu, Matt Klenk
Abstract: Understanding consumer preferences is essential to product design and predicting market response to these new products. Choice-based conjoint analysis is widely used to model user preferences using their choices in surveys. However, traditional conjoint estimation techniques assume simple linear models. This assumption may lead to limited predictability and inaccurate estimation of product attribute contributions, especially on data that has underlying non-linear relationships. In this work, we employ representation learning to efficiently alleviate this issue. We propose ConjointNet, which is composed of two novel neural architectures, to predict user preferences. We demonstrate that the proposed ConjointNet models outperform traditional conjoint estimate techniques on two preference datasets by over 5%, and offer insights into non-linear feature interactions.
Authors: Ehsan Latif, Xiaoming Zhai
Abstract: Data privacy remains a critical concern in educational research, necessitating Institutional Review Board (IRB) certification and stringent data handling protocols to ensure compliance with ethical standards. Traditional approaches rely on anonymization and controlled data-sharing mechanisms to facilitate research while mitigating privacy risks. However, these methods still involve direct access to raw student data, posing potential vulnerabilities and being time-consuming. This study proposes a federated learning (FL) framework for automatic scoring in educational assessments, eliminating the need to share raw data. Our approach leverages client-side model training, where student responses are processed locally on edge devices, and only optimized model parameters are shared with a central aggregation server. To effectively aggregate heterogeneous model updates, we introduce an adaptive weighted averaging strategy, which dynamically adjusts weight contributions based on client-specific learning characteristics. This method ensures robust model convergence while preserving privacy. We evaluate our framework using assessment data from nine middle schools, comparing the accuracy of federated learning-based scoring models with traditionally trained centralized models. A statistical significance test (paired t-test, $t(8) = 2.29, p = 0.051$) confirms that the accuracy difference between the two approaches is not statistically significant, demonstrating that federated learning achieves comparable performance while safeguarding student data. Furthermore, our method significantly reduces data collection, processing, and deployment overhead, accelerating the adoption of AI-driven educational assessments in a privacy-compliant manner.
Authors: Hanyang Zhao, Haoxian Chen, Yucheng Guo, Genta Indra Winata, Tingting Ou, Ziyu Huang, David D. Yao, Wenpin Tang
Abstract: We introduce Rich Preference Optimization (RPO), a novel pipeline that leverages rich feedback signals to improve the curation of preference pairs for fine-tuning text-to-image diffusion models. Traditional methods, like Diffusion-DPO, often rely solely on reward model labeling, which can be opaque, offer limited insights into the rationale behind preferences, and are prone to issues such as reward hacking or overfitting. In contrast, our approach begins with generating detailed critiques of synthesized images to extract reliable and actionable image editing instructions. By implementing these instructions, we create refined images, resulting in synthetic, informative preference pairs that serve as enhanced tuning datasets. We demonstrate the effectiveness of our pipeline and the resulting datasets in fine-tuning state-of-the-art diffusion models.
Authors: Hyunwoo Park, Baekryun Seong, Sang-Ki Ko
Abstract: In cooperative multi-agent reinforcement learning (MARL), the permutation problem where the state space grows exponentially with the number of agents reduces sample efficiency. Additionally, many existing architectures struggle with scalability, relying on a fixed structure tied to a specific number of agents, limiting their applicability to environments with a variable number of entities. While approaches such as graph neural networks (GNNs) and self-attention mechanisms have progressed in addressing these challenges, they have significant limitations as dense GNNs and self-attention mechanisms incur high computational costs. To overcome these limitations, we propose a novel agent network and a non-linear mixing network that ensure permutation-equivariance and scalability, allowing them to generalize to environments with various numbers of agents. Our agent network significantly reduces computational complexity, and our scalable hypernetwork enables efficient weight generation for non-linear mixing. Additionally, we introduce curriculum learning to improve training efficiency. Experiments on SMACv2 and Google Research Football (GRF) demonstrate that our approach achieves superior learning performance compared to existing methods. By addressing both permutation-invariance and scalability in MARL, our work provides a more efficient and adaptable framework for cooperative MARL. Our code is available at https://github.com/funny-rl/SPECTra.
Authors: Zekai Zhang, Dan Li, Shunyu Wu, Junya Cai, Bo Zhang, See Kiong Ng, Zibin Zheng
Abstract: Prognostic and Health Management (PHM) are crucial ways to avoid unnecessary maintenance for Cyber-Physical Systems (CPS) and improve system reliability. Predicting the Remaining Useful Life (RUL) is one of the most challenging tasks for PHM. Existing methods require prior knowledge about the system, contrived assumptions, or temporal mining to model the life cycles of machine equipment/devices, resulting in diminished accuracy and limited applicability in real-world scenarios. This paper proposes a Bi-directional Adversarial network with Covariate Encoding for machine Remaining Useful Life (BACE-RUL) prediction, which only adopts sensor measurements from the current life cycle to predict RUL rather than relying on previous consecutive cycle recordings. The current sensor measurements of mechanical devices are encoded to a conditional space to better understand the implicit inner mechanical status. The predictor is trained as a conditional generative network with the encoded sensor measurements as its conditions. Various experiments on several real-world datasets, including the turbofan aircraft engine dataset and the dataset collected from degradation experiments of Li-Ion battery cells, show that the proposed model is a general framework and outperforms state-of-the-art methods.
Authors: Andrew Starkey, Uduak Idio Akpan, Omaimah AL Hosni, Yaseen Pullissery
Abstract: There have been several attempts to develop Feature Selection (FS) algorithms capable of identifying features that are relevant in a dataset. Although in certain applications the FS algorithms can be seen to be successful, they have similar basic limitations. In all cases, the global feature selection algorithms seek to select features that are relevant and common to all classes of the dataset. This is a major limitation since there could be features that are specifically useful for a particular class while irrelevant for other classes, and full explanation of the relationship at class level therefore cannot be determined. While the inclusion of such features for all classes could cause improved predictive ability for the relevant class, the same features could be problematic for other classes. In this paper, we examine this issue and also develop a class-level feature selection method called the Feature Weighted Growing Self-Organising Map (FWGSOM). The proposed method carries out feature analysis at class level which enhances its ability to identify relevant features for each class. Results from experiments indicate that our method performs better than other methods, gives explainable results at class level, and has a low computational footprint when compared to other methods.
Authors: Jiseong Park, Hanjin Kim, Seojin Kim, Jueun Choi
Abstract: Graph pooling, which compresses a whole graph into a smaller coarsened graph, is an essential component of graph representation learning. To efficiently compress a given graph, graph pooling methods often drop their nodes with attention-based scoring with the task loss. However, this often results in simply removing nodes with lower degrees without consideration of their feature-level relevance to the given task. To fix this problem, we propose a Multi-View Pruning(MVP), a graph pruning method based on a multi-view framework and reconstruction loss. Given a graph, MVP first constructs multiple graphs for different views either by utilizing the predefined modalities or by randomly partitioning the input features, to consider the importance of each node in diverse perspectives. Then, it learns the score for each node by considering both the reconstruction and the task loss. MVP can be incorporated with any hierarchical pooling framework to score the nodes. We validate MVP on multiple benchmark datasets by coupling it with two graph pooling methods, and show that it significantly improves the performance of the base graph pooling method, outperforming all baselines. Further analysis shows that both the encoding of multiple views and the consideration of reconstruction loss are the key to the success of MVP, and that it indeed identifies nodes that are less important according to domain knowledge.
Authors: Zirui Yuan, Siqi Lai, Hao Liu
Abstract: Traffic Signal Control (TSC) plays a critical role in urban traffic management by optimizing traffic flow and mitigating congestion. While Large Language Models (LLMs) have recently emerged as promising tools for TSC due to their exceptional problem-solving and generalization capabilities, existing approaches fail to address the essential need for inter-agent coordination, limiting their effectiveness in achieving network-wide optimization. To bridge this gap, we propose CoLLMLight, a cooperative LLM agent framework for TSC. Specifically, we first construct a structured spatiotemporal graph to capture real-time traffic dynamics and spatial relationships among neighboring intersections, enabling the LLM to reason about complex traffic interactions. Moreover, we introduce a complexity-aware reasoning mechanism that dynamically adapts reasoning depth based on real-time traffic conditions, ensuring optimal computational efficiency without sacrificing decision quality. Besides, we propose a fine-tuning strategy that leverages iterative simulation-driven data collection and environmental feedback to build a lightweight LLM tailored for cooperative TSC. Extensive experiments on both synthetic and real-world datasets demonstrate that CoLLMLight outperforms state-of-the-art methods in diverse traffic scenarios, showcasing its effectiveness, scalability, and robustness.
Authors: Jian Qian, Teck Lun Goh, Bingyu Xie, Chengyao Zhu, Biao Wan, Yawen Guan, Patrick Yin Chiang
Abstract: Biological signals, such as electroencephalograms (EEGs) and electrocardiograms (ECGs), play a pivotal role in numerous clinical practices, such as diagnosing brain and cardiac arrhythmic diseases. Existing methods for biosignal classification rely on Attention-based frameworks with dense Feed Forward layers, which lead to inefficient learning, high computational overhead, and suboptimal performance. In this work, we introduce BioMamba, a Spectro-Temporal Embedding strategy applied to the Bidirectional Mamba framework with Sparse Feed Forward layers to enable effective learning of biosignal sequences. By integrating these three key components, BioMamba effectively addresses the limitations of existing methods. Extensive experiments demonstrate that BioMamba significantly outperforms state-of-the-art methods with marked improvement in classification performance. The advantages of the proposed BioMamba include (1) Reliability: BioMamba consistently delivers robust results, confirmed across six evaluation metrics. (2) Efficiency: We assess both model and training efficiency, the BioMamba demonstrates computational effectiveness by reducing model size and resource consumption compared to existing approaches. (3) Generality: With the capacity to effectively classify a diverse set of tasks, BioMamba demonstrates adaptability and effectiveness across various domains and applications.
Authors: Zhixuan Lian, Shangyu Li, Qixuan Huang, Zijian Huang, Haifei Liu, Jianan Qiu, Puyu Yang, Laifa Tao
Abstract: Fault diagnosis of mechanical equipment involves data collection, feature extraction, and pattern recognition but is often hindered by the imbalanced nature of industrial data, introducing significant uncertainty and reducing diagnostic reliability. To address these challenges, this study proposes the Uncertainty-Aware Bayesian Meta-Learning Framework (UBMF), which integrates four key modules: data perturbation injection for enhancing feature robustness, cross-task self-supervised feature extraction for improving transferability, uncertainty-based sample filtering for robust out-of-domain generalization, and Bayesian meta-knowledge integration for fine-grained classification. Experimental results on ten open-source datasets under various imbalanced conditions, including cross-task, small-sample, and unseen-sample scenarios, demonstrate the superiority of UBMF, achieving an average improvement of 42.22% across ten Any-way 1-5-shot diagnostic tasks. This integrated framework effectively enhances diagnostic accuracy, generalization, and adaptability, providing a reliable solution for complex industrial fault diagnosis.
Authors: Jun-Gi Jang, Jingrui He, Andrew Margenot, Hanghang Tong
Abstract: Many real-world data, such as recommendation data and temporal graphs, can be represented as incomplete sparse tensors where most entries are unobserved. For such sparse tensors, identifying the top-k higher-order interactions that are most likely to occur among unobserved ones is crucial. Tensor factorization (TF) has gained significant attention in various tensor-based applications, serving as an effective method for finding these top-k potential interactions. However, existing TF methods primarily focus on effectively fusing latent vectors of entities, which limits their expressiveness. Since most entities in sparse tensors have only a few interactions, their latent representations are often insufficiently trained. In this paper, we propose TCN, an accurate and compatible tensor convolutional network that integrates seamlessly with existing TF methods for predicting higher-order interactions. We design a highly effective encoder to generate expressive latent vectors of entities. To achieve this, we propose to (1) construct a graph structure derived from a sparse tensor and (2) develop a relation-aware encoder, TCN, that learns latent representations of entities by leveraging the graph structure. Since TCN complements traditional TF methods, we seamlessly integrate TCN with existing TF methods, enhancing the performance of predicting top-k interactions. Extensive experiments show that TCN integrated with a TF method outperforms competitors, including TF methods and a hyperedge prediction method. Moreover, TCN is broadly compatible with various TF methods and GNNs (Graph Neural Networks), making it a versatile solution.
Authors: Alisa Sheinkman, Sara Wade
Abstract: As modern neural networks get more complex, specifying a model with high predictive performance and sound uncertainty quantification becomes a more challenging task. Despite some promising theoretical results on the true posterior predictive distribution of Bayesian neural networks, the properties of even the most commonly used posterior approximations are often questioned. Computational burdens and intractable posteriors expose miscalibrated Bayesian neural networks to poor accuracy and unreliable uncertainty estimates. Approximate Bayesian inference aims to replace unknown and intractable posterior distributions with some simpler but feasible distributions. The dimensions of modern deep models coupled with the lack of identifiability make Markov chain Monte Carlo tremendously expensive and unable to fully explore the multimodal posterior. On the other hand, variational inference benefits from improved computational complexity but lacks the asymptotical guarantees of sampling-based inference and tends to concentrate around a single mode. The performance of both approaches heavily depends on architectural choices; this paper aims to shed some light on this, by considering the computational costs, accuracy and uncertainty quantification in different scenarios including large width and out-of-sample data. To improve posterior exploration, different model averaging and ensembling techniques are studied, along with their benefits on predictive performance. In our experiments, variational inference overall provided better uncertainty quantification than Markov chain Monte Carlo; further, stacking and ensembles of variational approximations provided comparable to Markov chain Monte Carlo accuracy at a much-reduced cost.
Authors: Yigit Efe Erginbas, Thomas A. Courtade, Kannan Ramchandran
Abstract: We consider an assortment selection and pricing problem in which a seller has $N$ different items available for sale. In each round, the seller observes a $d$-dimensional contextual preference information vector for the user, and offers to the user an assortment of $K$ items at prices chosen by the seller. The user selects at most one of the products from the offered assortment according to a multinomial logit choice model whose parameters are unknown. The seller observes which, if any, item is chosen at the end of each round, with the goal of maximizing cumulative revenue over a selling horizon of length $T$. For this problem, we propose an algorithm that learns from user feedback and achieves a revenue regret of order $\widetilde{O}(d \sqrt{K T} / L_0 )$ where $L_0$ is the minimum price sensitivity parameter. We also obtain a lower bound of order $\Omega(d \sqrt{T}/ L_0)$ for the regret achievable by any algorithm.
Authors: Tuomas Jalonen, Mohammad Al-Sa'd, Serkan Kiranyaz, Moncef Gabbouj
Abstract: Neural networks require massive amounts of annotated data to train intelligent solutions. Acquiring many labeled data in industrial applications is often difficult; therefore, semi-supervised approaches are preferred. We propose a new semi-supervised co-training method, which combines time and time-frequency (TF) machine learning models to improve performance and reliability. The developed framework collaboratively co-trains fast time-domain models by utilizing high-performing TF techniques without increasing the inference complexity. Besides, it operates in cloud-edge networks and offers holistic support for many applications covering edge-real-time monitoring and cloud-based updates and corrections. Experimental results on bearing fault diagnosis verify the superiority of our technique compared to a competing self-training method. The results from two case studies show that our method outperforms self-training for different noise levels and amounts of available data with accuracy gains reaching from 10.6% to 33.9%. They demonstrate that fusing time-domain and TF-based models offers opportunities for developing high-performance industrial solutions.
Authors: Chengyan Jiang, Jiamin Fan, Talal Halabi, Israat Haque
Abstract: The widespread adoption of smartphones and smart wearable devices has led to the widespread use of Centralized Federated Learning (CFL) for training powerful machine learning models while preserving data privacy. However, CFL faces limitations due to its overreliance on a central server, which impacts latency and system robustness. Decentralized Federated Learning (DFL) is introduced to address these challenges. It facilitates direct collaboration among participating devices without relying on a central server. Each device can independently connect with other devices and share model parameters. This work explores crucial factors influencing the convergence and generalization capacity of DFL models, emphasizing network topologies, non-IID data distribution, and training strategies. We first derive the convergence rate of different DFL model deployment strategies. Then, we comprehensively analyze various network topologies (e.g., linear, ring, star, and mesh) with different degrees of non-IID data and evaluate them over widely adopted machine learning models (e.g., classical, deep neural networks, and Large Language Models) and real-world datasets. The results reveal that models converge to the optimal one for IID data. However, the convergence rate is inversely proportional to the degree of non-IID data distribution. Our findings will serve as valuable guidelines for designing effective DFL model deployments in practical applications.
Authors: Haoxin Liu, Harshavardhan Kamarthi, Zhiyuan Zhao, Shangqing Xu, Shiyu Wang, Qingsong Wen, Tom Hartvigsen, Fei Wang, B. Aditya Prakash
Abstract: Time series analysis (TSA) is a longstanding research topic in the data mining community and has wide real-world significance. Compared to "richer" modalities such as language and vision, which have recently experienced explosive development and are densely connected, the time-series modality remains relatively underexplored and isolated. We notice that many recent TSA works have formed a new research field, i.e., Multiple Modalities for TSA (MM4TSA). In general, these MM4TSA works follow a common motivation: how TSA can benefit from multiple modalities. This survey is the first to offer a comprehensive review and a detailed outlook for this emerging field. Specifically, we systematically discuss three benefits: (1) reusing foundation models of other modalities for efficient TSA, (2) multimodal extension for enhanced TSA, and (3) cross-modality interaction for advanced TSA. We further group the works by the introduced modality type, including text, images, audio, tables, and others, within each perspective. Finally, we identify the gaps with future opportunities, including the reused modalities selections, heterogeneous modality combinations, and unseen tasks generalizations, corresponding to the three benefits. We release an up-to-date GitHub repository that includes key papers and resources.
Authors: Halil Alperen Gozeten, M. Emrullah Ildiz, Xuechen Zhang, Mahdi Soltanolkotabi, Marco Mondelli, Samet Oymak
Abstract: Test-time training (TTT) methods explicitly update the weights of a model to adapt to the specific test instance, and they have found success in a variety of settings, including most recently language modeling and reasoning. To demystify this success, we investigate a gradient-based TTT algorithm for in-context learning, where we train a transformer model on the in-context demonstrations provided in the test prompt. Specifically, we provide a comprehensive theoretical characterization of linear transformers when the update rule is a single gradient step. Our theory (i) delineates the role of alignment between pretraining distribution and target task, (ii) demystifies how TTT can alleviate distribution shift, and (iii) quantifies the sample complexity of TTT including how it can significantly reduce the eventual sample size required for in-context learning. As our empirical contribution, we study the benefits of TTT for TabPFN, a tabular foundation model. In line with our theory, we demonstrate that TTT significantly reduces the required sample size for tabular classification (3 to 5 times fewer) unlocking substantial inference efficiency with a negligible training cost.
Authors: Maximiliane Rautenstrau{\ss}, Maximilian Schiffer
Abstract: Minimizing response times to meet legal requirements and serve patients in a timely manner is crucial for Emergency Medical Service (EMS) systems. Achieving this goal necessitates optimizing operational decision-making to efficiently manage ambulances. Against this background, we study a centrally controlled EMS system for which we learn an online ambulance dispatching and redeployment policy that aims at minimizing the mean response time of ambulances within the system by dispatching an ambulance upon receiving an emergency call and redeploying it to a waiting location upon the completion of its service. We propose a novel combinatorial optimization-augmented machine learning pipeline that allows to learn efficient policies for ambulance dispatching and redeployment. In this context, we further show how to solve the underlying full-information problem to generate training data and propose an augmentation scheme that improves our pipeline's generalization performance by mitigating a possible distribution mismatch with respect to the considered state space. Compared to existing methods that rely on augmentation during training, our approach offers substantial runtime savings of up to 87.9% while yielding competitive performance. To evaluate the performance of our pipeline against current industry practices, we conduct a numerical case study on the example of San Francisco's 911 call data. Results show that the learned policies outperform the online benchmarks across various resource and demand scenarios, yielding a reduction in mean response time of up to 30%.
Authors: Yekta Amirkhalili, Ho Yi Wong
Abstract: The rapid growth of mobile banking (m-banking), especially after the COVID-19 pandemic, has reshaped the financial sector. This study analyzes consumer reviews of m-banking apps from five major Canadian banks, collected from Google Play and iOS App stores. Sentiment analysis and topic modeling classify reviews as positive, neutral, or negative, highlighting user preferences and areas for improvement. Data pre-processing was performed with NLTK, a Python language processing tool, and topic modeling used Latent Dirichlet Allocation (LDA). Sentiment analysis compared methods, with Long Short-Term Memory (LSTM) achieving 82\% accuracy for iOS reviews and Multinomial Naive Bayes 77\% for Google Play. Positive reviews praised usability, reliability, and features, while negative reviews identified login issues, glitches, and dissatisfaction with updates.This is the first study to analyze both iOS and Google Play m-banking app reviews, offering insights into app strengths and weaknesses. Findings underscore the importance of user-friendly designs, stable updates, and better customer service. Advanced text analytics provide actionable recommendations for improving user satisfaction and experience.
Authors: Jieming Bian, Lei Wang, Letian Zhang, Jie Xu
Abstract: Fine-tuning large language models (LLMs) in federated settings enables privacy-preserving adaptation but suffers from cross-client interference due to model aggregation. Existing federated LoRA fine-tuning methods, primarily based on FedAvg, struggle with data heterogeneity, leading to harmful cross-client interference and suboptimal personalization. In this work, we propose \textbf{FedALT}, a novel personalized federated LoRA fine-tuning algorithm that fundamentally departs from FedAvg. Instead of using an aggregated model to initialize local training, each client continues training its individual LoRA while incorporating shared knowledge through a separate Rest-of-the-World (RoTW) LoRA component. To effectively balance local adaptation and global information, FedALT introduces an adaptive mixer that dynamically learns input-specific weightings between the individual and RoTW LoRA components using the Mixture-of-Experts (MoE) principle. Through extensive experiments on NLP benchmarks, we demonstrate that FedALT significantly outperforms state-of-the-art personalized federated LoRA fine-tuning methods, achieving superior local adaptation without sacrificing computational efficiency.
Authors: Gabriel Clara, Sophie Langer, Johannes Schmidt-Hieber
Abstract: We analyze the landscape and training dynamics of diagonal linear networks in a linear regression task, with the network parameters being perturbed by small isotropic normal noise. The addition of such noise may be interpreted as a stochastic form of sharpness-aware minimization (SAM) and we prove several results that relate its action on the underlying landscape and training dynamics to the sharpness of the loss. In particular, the noise changes the expected gradient to force balancing of the weight matrices at a fast rate along the descent trajectory. In the diagonal linear model, we show that this equates to minimizing the average sharpness, as well as the trace of the Hessian matrix, among all possible factorizations of the same matrix. Further, the noise forces the gradient descent iterates towards a shrinkage-thresholding of the underlying true parameter, with the noise level explicitly regulating both the shrinkage factor and the threshold.
Authors: Da Long, Shandian Zhe, Samuel Williams, Leonid Oliker, Zhe Bai
Abstract: Simulating the long-term dynamics of multi-scale and multi-physics systems poses a significant challenge in understanding complex phenomena across science and engineering. The complexity arises from the intricate interactions between scales and the interplay of diverse physical processes. Neural operators have emerged as promising models for predicting such dynamics due to their flexibility and computational efficiency. However, they often fail to effectively capture multi-scale interactions or quantify the uncertainties inherent in the predictions. These limitations lead to rapid error accumulation, particularly in long-term forecasting of systems characterized by complex and coupled dynamics. To address these challenges, we propose a spatio-temporal Fourier transformer (StFT), in which each transformer block is designed to learn dynamics at a specific scale. By leveraging a structured hierarchy of StFT blocks, the model explicitly captures dynamics across both macro- and micro- spatial scales. Furthermore, a generative residual correction mechanism is integrated to estimate and mitigate predictive uncertainties, enhancing both the accuracy and reliability of long-term forecasts. Evaluations conducted on three benchmark datasets (plasma, fluid, and atmospheric dynamics) demonstrate the advantages of our approach over state-of-the-art ML methods.
Authors: Lauren Harrell, Christine Kaeser-Chen, Burcu Karagol Ayan, Keith Anderson, Michelangelo Conserva, Elise Kleeman, Maxim Neumann, Matt Overlan, Melissa Chapman, Drew Purves
Abstract: Species distribution models (SDMs) are necessary for measuring and predicting occurrences and habitat suitability of species and their relationship with environmental factors. We introduce a novel presence-only SDM with graph neural networks (GNN). In our model, species and locations are treated as two distinct node sets, and the learning task is predicting detection records as the edges that connect locations to species. Using GNN for SDM allows us to model fine-grained interactions between species and the environment. We evaluate the potential of this methodology on the six-region dataset compiled by National Center for Ecological Analysis and Synthesis (NCEAS) for benchmarking SDMs. For each of the regions, the heterogeneous GNN model is comparable to or outperforms previously-benchmarked single-species SDMs as well as a feed-forward neural network baseline model.
Authors: Eduard Tulchinskii, Daria Voronkova, Ilya Trofimov, Evgeny Burnaev, Serguei Barannikov
Abstract: Topological methods for comparing weighted graphs are valuable in various learning tasks but often suffer from computational inefficiency on large datasets. We introduce RTD-Lite, a scalable algorithm that efficiently compares topological features, specifically connectivity or cluster structures at arbitrary scales, of two weighted graphs with one-to-one correspondence between vertices. Using minimal spanning trees in auxiliary graphs, RTD-Lite captures topological discrepancies with $O(n^2)$ time and memory complexity. This efficiency enables its application in tasks like dimensionality reduction and neural network training. Experiments on synthetic and real-world datasets demonstrate that RTD-Lite effectively identifies topological differences while significantly reducing computation time compared to existing methods. Moreover, integrating RTD-Lite into neural network training as a loss function component enhances the preservation of topological structures in learned representations. Our code is publicly available at https://github.com/ArGintum/RTD-Lite
Authors: Atoosa Malemir Chegini, Keivan Rezaei, Hamid Eghbalzadeh, Soheil Feizi
Abstract: Fact-checking tabular data is essential for ensuring the accuracy of structured information. However, existing methods often rely on black-box models with opaque reasoning. We introduce RePanda, a structured fact verification approach that translates claims into executable pandas queries, enabling interpretable and verifiable reasoning. To train RePanda, we construct PanTabFact, a structured dataset derived from the TabFact train set, where claims are paired with executable queries generated using DeepSeek-Chat and refined through automated error correction. Fine-tuning DeepSeek-coder-7B-instruct-v1.5 on PanTabFact, RePanda achieves 84.09% accuracy on the TabFact test set. To evaluate Out-of-Distribution (OOD) generalization, we interpret question-answer pairs from WikiTableQuestions as factual claims and refer to this dataset as WikiFact. Without additional fine-tuning, RePanda achieves 84.72% accuracy on WikiFact, significantly outperforming all other baselines and demonstrating strong OOD robustness. Notably, these results closely match the zero-shot performance of DeepSeek-Chat (671B), indicating that our fine-tuning approach effectively distills structured reasoning from a much larger model into a compact, locally executable 7B model. Beyond fact verification, RePanda extends to tabular question answering by generating executable queries that retrieve precise answers. To support this, we introduce PanWiki, a dataset mapping WikiTableQuestions to pandas queries. Fine-tuning on PanWiki, RePanda achieves 75.1% accuracy in direct answer retrieval. These results highlight the effectiveness of structured execution-based reasoning for tabular verification and question answering. We have publicly released the dataset on Hugging Face at datasets/AtoosaChegini/PanTabFact.
Authors: Patricia Medina
Abstract: In this paper, we address the enhancement of classification accuracy for 3D point cloud Lidar data, an optical remote sensing technique that estimates the three-dimensional coordinates of a given terrain. Our approach introduces product coefficients, theoretical quantities derived from measure theory, as additional features in the classification process. We define and present the formulation of these product coefficients and conduct a comparative study, using them alongside principal component analysis (PCA) as feature inputs. Results demonstrate that incorporating product coefficients into the feature set significantly improves classification accuracy within this new framework.
Authors: Zhihao Zeng, Ziquan Fang, Yuting Huang, Lu Chen, Yunjun Gao
Abstract: Traffic prediction targets forecasting future traffic conditions using historical traffic data, serving a critical role in urban computing and transportation management. To mitigate the scarcity of traffic data while maintaining data privacy, numerous Federated Traffic Knowledge Transfer (FTT) approaches have been developed, which use transfer learning and federated learning to transfer traffic knowledge from data-rich cities to data-scarce cities, enhancing traffic prediction capabilities for the latter. However, current FTT approaches face challenges such as privacy leakage, cross-city data distribution discrepancies, low data quality, and inefficient knowledge transfer, limiting their privacy protection, effectiveness, robustness, and efficiency in real-world applications. To this end, we propose FedTT, an effective, efficient, and privacy-aware cross-city traffic knowledge transfer framework that transforms the traffic data domain from the data-rich cities and trains traffic models using the transformed data for the data-scarce cities. First, to safeguard data privacy, we propose a traffic secret transmission method that securely transmits and aggregates traffic domain-transformed data from source cities using a lightweight secret aggregation approach. Second, to mitigate the impact of traffic data distribution discrepancies on model performance, we introduce a traffic domain adapter to uniformly transform traffic data from the source cities' domains to that of the target city. Third, to improve traffic data quality, we design a traffic view imputation method to fill in and predict missing traffic data. Finally, to enhance transfer efficiency, FedTT is equipped with a federated parallel training method that enables the simultaneous training of multiple modules. Extensive experiments using 4 real-life datasets demonstrate that FedTT outperforms the 14 state-of-the-art baselines.
Authors: Jasmeet Kaur
Abstract: Effective uncertainty quantification is important for training modern predictive models with limited data, enhancing both accuracy and robustness. While Bayesian methods are effective for this purpose, they can be challenging to scale. When employing approximate Bayesian inference, ensuring the quality of samples from the posterior distribution in a computationally efficient manner is essential. This paper addresses the estimation of the Bayesian posterior to generate diverse samples by approximating the gradient flow of the Kullback-Leibler (KL) divergence and the cross entropy of the target approximation under the metric induced by the Stein Operator. It presents empirical evaluations on classification tasks to assess the method's performance and discuss its effectiveness for Model-Based Reinforcement Learning that uses uncertainty-aware network dynamics models.
Authors: Xi Wang, Hideaki Shimazaki
Abstract: We introduce a novel framework for learning in neural networks by decomposing each neuron's weight vector into two distinct parts, $W_1$ and $W_2$, thereby modeling contrastive information directly at the neuron level. Traditional gradient descent stores both positive (target) and negative (non-target) feature information in a single weight vector, often obscuring fine-grained distinctions. Our approach, by contrast, maintains separate updates for target and non-target features, ultimately forming a single effective weight $W = W_1 - W_2$ that is more robust to noise and class imbalance. Experimental results on both regression (California Housing, Wine Quality) and classification (MNIST, Fashion-MNIST, CIFAR-10) tasks suggest that this decomposition enhances generalization and resists overfitting, especially when training data are sparse or noisy. Crucially, the inference complexity remains the same as in the standard $WX + \text{bias}$ setup, offering a practical solution for improved learning without additional inference-time overhead.
Authors: Haonan Pan, Shuheng Chen, Elham Pishgar, Kamiar Alaei, Greg Placencia, Maryam Pishgar
Abstract: Coronary artery disease remains one of the leading causes of mortality globally. Despite advances in revascularization treatments like PCI and CABG, postoperative stroke is inevitable. This study aims to develop and evaluate a sophisticated machine learning prediction model to assess postoperative stroke risk in coronary revascularization patients.This research employed data from the MIMIC-IV database, consisting of a cohort of 7023 individuals. Study data included clinical, laboratory, and comorbidity variables. To reduce multicollinearity, variables with over 30% missing values and features with a correlation coefficient larger than 0.9 were deleted. The dataset has 70% training and 30% test. The Random Forest technique interpolated residual dataset missing values. Numerical values were normalized, whereas categorical variables were one-hot encoded. LASSO regularization selected features, and grid search found model hyperparameters. Finally, Logistic Regression, XGBoost, SVM, and CatBoost were employed for predictive modeling, and SHAP analysis assessed stroke risk for each variable. AUC of 0.855 (0.829-0.878) showed that SVM model outperformed logistic regression and CatBoost models in prior research. SHAP research showed that the Charlson Comorbidity Index (CCI), diabetes, chronic kidney disease, and heart failure are significant prognostic factors for postoperative stroke. This study shows that improved machine learning reduces overfitting and improves model predictive accuracy. Models using the CCI alone cannot predict postoperative stroke risk as accurately as those using independent comorbidity variables. The suggested technique provides a more thorough and individualized risk assessment by encompassing a wider range of clinically relevant characteristics, making it a better reference for preoperative risk assessments and targeted intervention.
Authors: Sumana Sanyasipura Nagaraju
Abstract: Effective feature selection, representation and transformation are principal steps in machine learning to improve prediction accuracy, model generalization and computational efficiency. Reinforcement learning provides a new perspective towards balanced exploration of optimal feature subset using multi-agent and single-agent models. Interactive reinforcement learning integrated with decision tree improves feature knowledge, state representation and selection efficiency, while diversified teaching strategies improve both selection quality and efficiency. The state representation can further be enhanced by scanning features sequentially along with the usage of convolutional auto-encoder. Monte Carlo-based reinforced feature selection(MCRFS), a single-agent feature selection method reduces computational burden by incorporating early-stopping and reward-level interactive strategies. A dual-agent RL framework is also introduced that collectively selects features and instances, capturing the interactions between them. This enables the agents to navigate through complex data spaces. To outperform the traditional feature engineering, cascading reinforced agents are used to iteratively improve the feature space, which is a self-optimizing framework. The blend of reinforcement learning, multi-agent systems, and bandit-based approaches offers exciting paths for studying scalable and interpretable machine learning solutions to handle high-dimensional data and challenging predictive tasks.
Authors: Xiaoyu Wu, Yifei Pang, Terrance Liu, Steven Wu
Abstract: Tabular data synthesis using diffusion models has gained significant attention for its potential to balance data utility and privacy. However, existing privacy evaluations often rely on heuristic metrics or weak membership inference attacks (MIA), leaving privacy risks inadequately assessed. In this work, we conduct a rigorous MIA study on diffusion-based tabular synthesis, revealing that state-of-the-art attacks designed for image models fail in this setting. We identify noise initialization as a key factor influencing attack efficacy and propose a machine-learning-driven approach that leverages loss features across different noises and time steps. Our method, implemented with a lightweight MLP, effectively learns membership signals, eliminating the need for manual optimization. Experimental results from the MIDST Challenge @ SaTML 2025 demonstrate the effectiveness of our approach, securing first place across all tracks. Code is available at https://github.com/Nicholas0228/Tartan_Federer_MIDST.
Authors: Qingshi Sun, Nathan Justin, Andres Gomez, Phebe Vayanos
Abstract: Logistic regression models are widely used in the social and behavioral sciences and in high-stakes domains, due to their simplicity and interpretability properties. At the same time, such domains are permeated by distribution shifts, where the distribution generating the data changes between training and deployment. In this paper, we study a distributionally robust logistic regression problem that seeks the model that will perform best against adversarial realizations of the data distribution drawn from a suitably constructed Wasserstein ambiguity set. Our model and solution approach differ from prior work in that we can capture settings where the likelihood of distribution shifts can vary across features, significantly broadening the applicability of our model relative to the state-of-the-art. We propose a graph-based solution approach that can be integrated into off-the-shelf optimization solvers. We evaluate the performance of our model and algorithms on numerous publicly available datasets. Our solution achieves a 408x speed-up relative to the state-of-the-art. Additionally, compared to the state-of-the-art, our model reduces average calibration error by up to 36.19% and worst-case calibration error by up to 41.70%, while increasing the average area under the ROC curve (AUC) by up to 18.02% and worst-case AUC by up to 48.37%.
Authors: Yebo Wu, Chunlin Tian, Jingguang Li, He Sun, Kahou Tam, Li Li, Chengzhong Xu
Abstract: Large Language Models (LLMs) have achieved remarkable success across a wide range of tasks, with fine-tuning playing a pivotal role in adapting them to specific downstream applications. Federated Learning (FL) offers a promising approach that enables collaborative model adaptation while ensuring data privacy, i.e., FedLLM. In this survey, we provide a systematic and thorough review of the integration of LLMs with FL. Specifically, we first trace the historical evolution of both LLMs and FL, while summarizing relevant prior surveys. We then present an in-depth analysis of the fundamental challenges encountered in deploying FedLLM. Following this, we conduct an extensive study of existing parameter-efficient fine-tuning (PEFT) methods and explore their applicability in FL. Furthermore, we introduce a comprehensive evaluation benchmark to rigorously assess FedLLM performance and discuss its diverse real-world applications across multiple domains. Finally, we identify critical open challenges and outline promising research directions to drive future advancements in FedLLM. We maintain an active \href{https://github.com/Clin0212/Awesome-Federated-LLM-Learning}{GitHub repository} tracking cutting-edge advancements. This survey serves as a foundational resource for researchers and practitioners, offering insights into the evolving landscape of federated fine-tuning for LLMs while guiding future innovations in privacy-preserving AI.
URLs: https://github.com/Clin0212/Awesome-Federated-LLM-Learning
Authors: Jiafan He, Quanquan Gu
Abstract: Variance-dependent regret bounds for linear contextual bandits, which improve upon the classical $\tilde{O}(d\sqrt{K})$ regret bound to $\tilde{O}(d\sqrt{\sum_{k=1}^K\sigma_k^2})$, where $d$ is the context dimension, $K$ is the number of rounds, and $\sigma^2_k$ is the noise variance in round $k$, has been widely studied in recent years. However, most existing works focus on the regret upper bounds instead of lower bounds. To our knowledge, the only lower bound is from Jia et al. (2024), which proved that for any eluder dimension $d_{\textbf{elu}}$ and total variance budget $\Lambda$, there exists an instance with $\sum_{k=1}^K\sigma_k^2\leq \Lambda$ for which any algorithm incurs a variance-dependent lower bound of $\Omega(\sqrt{d_{\textbf{elu}}\Lambda})$. However, this lower bound has a $\sqrt{d}$ gap with existing upper bounds. Moreover, it only considers a fixed total variance budget $\Lambda$ and does not apply to a general variance sequence $\{\sigma_1^2,\ldots,\sigma_K^2\}$. In this paper, to overcome the limitations of Jia et al. (2024), we consider the general variance sequence under two settings. For a prefixed sequence, where the entire variance sequence is revealed to the learner at the beginning of the learning process, we establish a variance-dependent lower bound of $\Omega(d \sqrt{\sum_{k=1}^K\sigma_k^2 }/\log K)$ for linear contextual bandits. For an adaptive sequence, where an adversary can generate the variance $\sigma_k^2$ in each round $k$ based on historical observations, we show that when the adversary must generate $\sigma_k^2$ before observing the decision set $\mathcal{D}_k$, a similar lower bound of $\Omega(d\sqrt{ \sum_{k=1}^K\sigma_k^2} /\log^6(dK))$ holds. In both settings, our results match the upper bounds of the SAVE algorithm (Zhao et al., 2023) up to logarithmic factors.
Authors: Hang Ni, Jindong Han, Nengjun Zhu, Hao Liu
Abstract: Graph Anomaly Detection (GAD) plays a vital role in various data mining applications such as e-commerce fraud prevention and malicious user detection. Recently, Graph Neural Network (GNN) based approach has demonstrated great effectiveness in GAD by first encoding graph data into low-dimensional representations and then identifying anomalies under the guidance of supervised or unsupervised signals. However, existing GNN-based approaches implicitly follow the homophily principle (i.e., the "like attracts like" phenomenon) and fail to learn discriminative embedding for anomalies that connect vast normal nodes. Moreover, such approaches identify anomalies in a unified global perspective but overlook diversified abnormal patterns conditioned on local graph context, leading to suboptimal performance. To overcome the aforementioned limitations, in this paper, we propose a Multi-hypersphere Heterophilic Graph Learning (MHetGL) framework for unsupervised GAD. Specifically, we first devise a Heterophilic Graph Encoding (HGE) module to learn distinguishable representations for potential anomalies by purifying and augmenting their neighborhood in a fully unsupervised manner. Then, we propose a Multi-Hypersphere Learning (MHL) module to enhance the detection capability for context-dependent anomalies by jointly incorporating critical patterns from both global and local perspectives. Extensive experiments on ten real-world datasets show that MHetGL outperforms 14 baselines. Our code is publicly available at https://github.com/KennyNH/MHetGL.
Authors: Yuhao Zhou, Yuxin Tian, Jindi Lv, Mingjia Shi, Yuanxi Li, Qing Ye, Shuhao Zhang, Jiancheng Lv
Abstract: In the realm of high-frequency data streams, achieving real-time learning within varying memory constraints is paramount. This paper presents Ferret, a comprehensive framework designed to enhance online accuracy of Online Continual Learning (OCL) algorithms while dynamically adapting to varying memory budgets. Ferret employs a fine-grained pipeline parallelism strategy combined with an iterative gradient compensation algorithm, ensuring seamless handling of high-frequency data with minimal latency, and effectively counteracting the challenge of stale gradients in parallel training. To adapt to varying memory budgets, its automated model partitioning and pipeline planning optimizes performance regardless of memory limitations. Extensive experiments across 20 benchmarks and 5 integrated OCL algorithms show Ferret's remarkable efficiency, achieving up to 3.7$\times$ lower memory overhead to reach the same online accuracy compared to competing methods. Furthermore, Ferret consistently outperforms these methods across diverse memory budgets, underscoring its superior adaptability. These findings position Ferret as a premier solution for efficient and adaptive OCL framework in real-time environments.
Authors: Yuetong Yu, Ruiyang Ge, Ilker Hacihaliloglu, Alexander Rauscher, Roger Tam, Sophia Frangou
Abstract: Background: Patient stratification in brain disorders remains a significant challenge, despite advances in machine learning and multimodal neuroimaging. Automated machine learning algorithms have been widely applied for identifying patient subtypes (biotypes), but results have been inconsistent across studies. These inconsistencies are often attributed to algorithmic limitations, yet an overlooked factor may be the statistical properties of the input data. This study investigates the contribution of data patterns on algorithm performance by leveraging synthetic brain morphometry data as an exemplar. Methods: Four widely used algorithms-SuStaIn, HYDRA, SmileGAN, and SurrealGAN were evaluated using multiple synthetic pseudo-patient datasets designed to include varying numbers and sizes of clusters and degrees of complexity of morphometric changes. Ground truth, representing predefined clusters, allowed for the evaluation of performance accuracy across algorithms and datasets. Results: SuStaIn failed to process datasets with more than 17 variables, highlighting computational inefficiencies. HYDRA was able to perform individual-level classification in multiple datasets with no clear pattern explaining failures. SmileGAN and SurrealGAN outperformed other algorithms in identifying variable-based disease patterns, but these patterns were not able to provide individual-level classification. Conclusions: Dataset characteristics significantly influence algorithm performance, often more than algorithmic design. The findings emphasize the need for rigorous validation using synthetic data before real-world application and highlight the limitations of current clustering approaches in capturing the heterogeneity of brain disorders. These insights extend beyond neuroimaging and have implications for machine learning applications in biomedical research.
Authors: Wuzhou Sun, Siyi Li, Qingxiang Zou, Zixing Liao
Abstract: In various game scenarios, selecting a fixed number of targets from multiple enemy units is an extremely challenging task. This difficulty stems from the complex relationship between the threat levels of enemy units and their feature characteristics, which complicates the design of rule-based evaluators. Moreover, traditional supervised learning methods face the challenge of lacking explicit labels during training when applied to this threat evaluation problem. In this study, we redefine the threat evaluation problem as a reinforcement learning task and introduce an efficient evaluator training algorithm, Eval-PPO, based on the Proximal Policy Optimization (PPO) algorithm. Eval-PPO integrates multidimensional enemy features and the state information of friendly units through systematic training, thereby achieving precise threat assessment. Compared with rule-based methods, Eval-PPO demonstrates a significant improvement in average success rate, with an increase of 17.84%.
Authors: Sebastian Pineda Arango, Pedro Mercado, Shubham Kapoor, Abdul Fatir Ansari, Lorenzo Stella, Huibin Shen, Hugo Senetaire, Caner Turkmen, Oleksandr Shchur, Danielle C. Maddix, Michael Bohlke-Schneider, Yuyang Wang, Syama Sundar Rangapuram
Abstract: Covariates provide valuable information on external factors that influence time series and are critical in many real-world time series forecasting tasks. For example, in retail, covariates may indicate promotions or peak dates such as holiday seasons that heavily influence demand forecasts. Recent advances in pretraining large language model architectures for time series forecasting have led to highly accurate forecasters. However, the majority of these models do not readily use covariates as they are often specific to a certain task or domain. This paper introduces a new method to incorporate covariates into pretrained time series forecasting models. Our proposed approach incorporates covariate information into pretrained forecasting models through modular blocks that inject past and future covariate information, without necessarily modifying the pretrained model in consideration. In order to evaluate our approach, we introduce a benchmark composed of 32 different synthetic datasets with varying dynamics to evaluate the effectivity of forecasting models with covariates. Extensive evaluations on both synthetic and real datasets show that our approach effectively incorporates covariate information into pretrained models, outperforming existing baselines.
Authors: Hun Kang, Kyoungok Kim
Abstract: Isolation Forest (iForest) is an unsupervised anomaly detection algorithm designed to effectively detect anomalies under the assumption that anomalies are ``few and different." Various studies have aimed to enhance iForest, but the resulting algorithms often exhibited significant performance disparities across datasets. Additionally, the challenge of isolating rare and widely distributed anomalies persisted in research focused on improving splits. To address these challenges, we introduce Robust iForest (RiForest). RiForest leverages both existing features and random hyperplanes obtained through soft sparse random projection to identify superior split features for anomaly detection, independent of datasets. It utilizes the underutilized valley emphasis method for optimal split point determination and incorporates sparsity randomization in soft sparse random projection for enhanced anomaly detection robustness. Across 24 benchmark datasets, experiments demonstrate RiForest's consistent outperformance of existing algorithms in anomaly detection, emphasizing stability and robustness to noise variables.
Authors: Ertu\u{g}rul Ke\c{c}eci, M\"ujde G\"uzelkaya, Tufan Kumbasar
Abstract: This paper presents FedAlign, a Federated Learning (FL) framework particularly designed for System Identification (SYSID) tasks by aligning state representations. Local workers can learn State-Space Models (SSMs) with equivalent representations but different dynamics. We demonstrate that directly aggregating these local SSMs via FedAvg results in a global model with altered system dynamics. FedAlign overcomes this problem by employing similarity transformation matrices to align state representations of local SSMs, thereby establishing a common parameter basin that retains the dynamics of local SSMs. FedAlign computes similarity transformation matrices via two distinct approaches: FedAlign-A and FedAlign-O. In FedAlign-A, we represent the global SSM in controllable canonical form (CCF). We apply control theory to analytically derive similarity transformation matrices that convert each local SSM into this form. Yet, establishing global SSM in CCF brings additional alignment challenges in multi input - multi output SYSID as CCF representation is not unique, unlike in single input - single output SYSID. In FedAlign-O, we address these alignment challenges by reformulating the local parameter basin alignment problem as an optimization task. We determine the parameter basin of a local worker as the common parameter basin and solve least square problems to obtain similarity transformation matrices needed to align the remaining local SSMs. Through the experiments conducted on synthetic and real-world datasets, we show that FedAlign outperforms FedAvg, converges faster, and provides improved stability of the global SSM thanks to the efficient alignment of local parameter basins.
Authors: Yunbo Long, Liming Xu, Alexandra Brintrup
Abstract: Link prediction is crucial for uncovering hidden connections within complex networks, enabling applications such as identifying potential customers and products. However, this research faces significant challenges, including concerns about data privacy, as well as high computational and storage costs, especially when dealing with large-scale networks. Condensed graphs, which are much smaller than the original graphs while retaining essential information, has become an effective solution to both maintain data utility and preserve privacy. Existing methods, however, initialize synthetic graphs through random node selection without considering node connectivity, and are mainly designed for node classification tasks. As a result, their potential for privacy-preserving link prediction remains largely unexplored. We introduce HyDRO\textsuperscript{+}, a graph condensation method guided by algebraic Jaccard similarity, which leverages local connectivity information to optimize condensed graph structures. Extensive experiments on four real-world networks show that our method outperforms state-of-the-art methods and even the original networks in balancing link prediction accuracy and privacy preservation. Moreover, our method achieves nearly 20* faster training and reduces storage requirements by 452*, as demonstrated on the Computers dataset, compared to link prediction on the original networks. This work represents the first attempt to leverage condensed graphs for privacy-preserving link prediction information sharing in real-world complex networks. It offers a promising pathway for preserving link prediction information while safeguarding privacy, advancing the use of graph condensation in large-scale networks with privacy concerns.
Authors: Tingting Wang, Jiaxin Su, Haobing Liu, Ruobing Jiang
Abstract: Node classification in graphs aims to predict the categories of unlabeled nodes by utilizing a small set of labeled nodes. However, weighted graphs often contain noisy edges and anomalous edge weights, which can distort fine-grained relationships between nodes and hinder accurate classification. We propose the Edge Weight-aware Graph Structure Learning (EWGSL) method, which combines weight learning and graph structure learning to address these issues. EWGSL improves node classification by redefining attention coefficients in graph attention networks to incorporate node features and edge weights. It also applies graph structure learning to sparsify attention coefficients and uses a modified InfoNCE loss function to enhance performance by adapting to denoised graph weights. Extensive experimental results show that EWGSL has an average Micro-F1 improvement of 17.8% compared with the best baseline.
Authors: Milan Pape\v{z}, Martin Rektoris, V\'aclav \v{S}m\'idl, Tom\'a\v{s} Pevn\'y
Abstract: Deep generative models (DGMs) have recently demonstrated remarkable success in capturing complex probability distributions over graphs. Although their excellent performance is attributed to powerful and scalable deep neural networks, it is, at the same time, exactly the presence of these highly non-linear transformations that makes DGMs intractable. Indeed, despite representing probability distributions, intractable DGMs deny probabilistic foundations by their inability to answer even the most basic inference queries without approximations or design choices specific to a very narrow range of queries. To address this limitation, we propose probabilistic graph circuits (PGCs), a framework of tractable DGMs that provide exact and efficient probabilistic inference over (arbitrary parts of) graphs. Nonetheless, achieving both exactness and efficiency is challenging in the permutation-invariant setting of graphs. We design PGCs that are inherently invariant and satisfy these two requirements, yet at the cost of low expressive power. Therefore, we investigate two alternative strategies to achieve the invariance: the first sacrifices the efficiency, and the second sacrifices the exactness. We demonstrate that ignoring the permutation invariance can have severe consequences in anomaly detection, and that the latter approach is competitive with, and sometimes better than, existing intractable DGMs in the context of molecular graph generation.
Authors: Kasra Arabi, R. Teal Witter, Chinmay Hegde, Niv Cohen
Abstract: Generative models have rapidly evolved to generate realistic outputs. However, their synthetic outputs increasingly challenge the clear distinction between natural and AI-generated content, necessitating robust watermarking techniques. Watermarks are typically expected to preserve the integrity of the target image, withstand removal attempts, and prevent unauthorized replication onto unrelated images. To address this need, recent methods embed persistent watermarks into images produced by diffusion models using the initial noise. Yet, to do so, they either distort the distribution of generated images or rely on searching through a long dictionary of used keys for detection. In this paper, we propose a novel watermarking method that embeds semantic information about the generated image directly into the watermark, enabling a distortion-free watermark that can be verified without requiring a database of key patterns. Instead, the key pattern can be inferred from the semantic embedding of the image using locality-sensitive hashing. Furthermore, conditioning the watermark detection on the original image content improves robustness against forgery attacks. To demonstrate that, we consider two largely overlooked attack strategies: (i) an attacker extracting the initial noise and generating a novel image with the same pattern; (ii) an attacker inserting an unrelated (potentially harmful) object into a watermarked image, possibly while preserving the watermark. We empirically validate our method's increased robustness to these attacks. Taken together, our results suggest that content-aware watermarks can mitigate risks arising from image-generative models.
Authors: Nir Ailon, Akhiad Bercovich, Omri Weinstein
Abstract: We propose a cheaper alternative bilinear operator to matrix-multiplication in deep neural networks (DNNs). Unlike many stubborn attempts to accelerate MatMuls in DNN inference, this operator is supported by capabilities of existing GPU hardware, most notably NVIDIA TensorCores. To our knowledge, this is the first GPU-native acceleration technique which \emph{does not decrease} (in fact, increases) the number of trainable parameters of the network, mitigating the accuracy-loss of compression-based techniques. Hence, this operator is at the same time more expressive than MatMul, yet requires substantially \emph{fewer} FLOPs to evaluate. We term this new operator \emph{Strassen-Tile} (STL). The main idea behind STL$(X,W)$ is a \emph{local} change-of-basis (learnable encoder) on weights and activation \emph{tiles}, after which we perform batched \emph{elementwise} products between tiles, and a final decoding transformation (inspired by algebraic pipelines from fast matrix and polynomial multiplication). We compare STL against two benchmarks. The first one is SoTA T2T-ViT on Imagenet-1K. Here we show that replacing \emph{all} linear layers with STL and training from scratch, results in factor x2.7 reduction in FLOPs with a 0.5 \emph{accuracy improvement}. Our second speed-accuracy comparison benchmark for pretrained LLMs is the most practical GPU-acceleration technique, \twofour structured Sparsity. Finetuning TinyLlama \cite{tinyllama24} with STL layers on the Slim Pajama dataset, achieves similar accuracy to 2:4, with x2.2 FLOP speedup compared to x1.7 of the latter. Finally, we discuss a group-theoretic approach for discovering \emph{universal} encoders for STL, which could lead to fast \emph{black-box} acceleration via approximate matrix-multiplication (AMM).
Authors: Sharmita Dey, Sarath Ravindran Nair
Abstract: We present a mutually aligned diffusion framework for cross-modal biomechanical motion generation, guided by a dynamical systems perspective. By treating each modality, e.g., observed joint angles ($X$) and ground reaction forces ($Y$), as complementary observations of a shared underlying locomotor dynamical system, our method aligns latent representations at each diffusion step, so that one modality can help denoise and disambiguate the other. Our alignment approach is motivated by the fact that local time windows of $X$ and $Y$ represent the same phase of an underlying dynamical system, thereby benefiting from a shared latent manifold. We introduce a simple local latent manifold alignment (LLMA) strategy that incorporates first-order and second-order alignment within the latent space for robust cross-modal biomechanical generation without bells and whistles. Through experiments on multimodal human biomechanics data, we show that aligning local latent dynamics across modalities improves generation fidelity and yields better representations.
Authors: Yunbo Long, Liming Xu, Ge Zheng, Alexandra Brintrup
Abstract: Federated learning (FL) enables retailers to share model parameters for demand forecasting while maintaining privacy. However, heterogeneous data across diverse regions, driven by factors such as varying consumer behavior, poses challenges to the effectiveness of federated learning. To tackle this challenge, we propose Bubble-Cluster Federated Learning (BFL), a novel clustering-based federated learning framework tailored for sales prediction. By leveraging differential privacy and feature importance distribution, BFL groups retailers into distinct "bubbles", each forming its own federated learning (FL) system to effectively isolate data heterogeneity. Within each bubble, Transformer models are designed to predict local sales for each client. Our experiments demonstrate that BFL significantly surpasses FedAvg and outperforms local learning in demand forecasting performance across all participating clients. Compared to local learning, BFL can achieve a 5.4\% improvement in R\textsuperscript{2}, a 69\% reduction in RMSE, and a 45\% decrease in MAE. Our study highlights BFL's adaptability in enabling effective federated learning through dynamic adjustments to noise levels and the range of clients participating in each bubble. This approach strategically groups participants into distinct "bubbles" while proactively identifying and filtering out risky clients that could compromise the FL system. The findings demonstrate BFL's ability to enhance collaborative learning in regression tasks on heterogeneous data, achieving a balance between forecasting accuracy and privacy preservation in retail applications. Additionally, BFL's capability to detect and neutralize poisoned data from clients enhances the system's robustness and reliability, ensuring more secure and effective federated learning.
Authors: Natinael Solomon Neggatu, Jeremie Houssineau, Giovanni Montana
Abstract: Offline reinforcement learning (RL) looks at learning how to optimally solve tasks using a fixed dataset of interactions from the environment. Many off-policy algorithms developed for online learning struggle in the offline setting as they tend to over-estimate the behaviour of out of distributions actions. Existing offline RL algorithms adapt off-policy algorithms, employing techniques such as constraining the policy or modifying the value function to achieve good performance on individual datasets but struggle to adapt to different tasks or datasets of different qualities without tuning hyper-parameters. We introduce a policy switching technique that dynamically combines the behaviour of a pure off-policy RL agent, for improving behaviour, and a behavioural cloning (BC) agent, for staying close to the data. We achieve this by using a combination of epistemic uncertainty, quantified by our RL model, and a metric for aleatoric uncertainty extracted from the dataset. We show empirically that our policy switching technique can outperform not only the individual algorithms used in the switching process but also compete with state-of-the-art methods on numerous benchmarks. Our use of epistemic uncertainty for policy switching also allows us to naturally extend our method to the domain of offline to online fine-tuning allowing our model to adapt quickly and safely from online data, either matching or exceeding the performance of current methods that typically require additional modification or hyper-parameter fine-tuning.
Authors: Soufiane Bacha, Huansheng Ning, Belarbi Mostefa, Doreen Sebastian Sarwatt, Sahraoui Dhelim
Abstract: Accurate illness diagnosis is vital for effective treatment and patient safety. Machine learning models are widely used for cancer diagnosis based on historical medical data. However, data imbalance remains a major challenge, leading to hindering classifier performance and reliability. The SMOTEBoost method addresses this issue by generating synthetic data to balance the dataset, but it may overlook crucial overlapping regions near the decision boundary and can produce noisy samples. This paper proposes RE-SMOTEBoost, an enhanced version of SMOTEBoost, designed to overcome these limitations. Firstly, RE-SMOTEBoost focuses on generating synthetic samples in overlapping regions to better capture the decision boundary using roulette wheel selection. Secondly, it incorporates a filtering mechanism based on information entropy to reduce noise, and borderline cases and improve the quality of generated data. Thirdly, we introduce a double regularization penalty to control the synthetic samples proximity to the decision boundary and avoid class overlap. These enhancements enable higher-quality oversampling of the minority class, resulting in a more balanced and effective training dataset. The proposed method outperforms existing state-of-the-art techniques when evaluated on imbalanced datasets. Compared to the top-performing sampling algorithms, RE-SMOTEBoost demonstrates a notable improvement of 3.22\% in accuracy and a variance reduction of 88.8\%. These results indicate that the proposed model offers a solid solution for medical settings, effectively overcoming data scarcity and severe imbalance caused by limited samples, data collection difficulties, and privacy constraints.
Authors: Liying Han, Gaofeng Dong, Xiaomin Ouyang, Lance Kaplan, Federico Cerutti, Mani Srivastava
Abstract: Complex events (CEs) play a crucial role in CPS-IoT applications, enabling high-level decision-making in domains such as smart monitoring and autonomous systems. However, most existing models focus on short-span perception tasks, lacking the long-term reasoning required for CE detection. CEs consist of sequences of short-time atomic events (AEs) governed by spatiotemporal dependencies. Detecting them is difficult due to long, noisy sensor data and the challenge of filtering out irrelevant AEs while capturing meaningful patterns. This work explores CE detection as a case study for CPS-IoT foundation models capable of long-term reasoning. We evaluate three approaches: (1) leveraging large language models (LLMs), (2) employing various neural architectures that learn CE rules from data, and (3) adopting a neurosymbolic approach that integrates neural models with symbolic engines embedding human knowledge. Our results show that the state-space model, Mamba, which belongs to the second category, outperforms all methods in accuracy and generalization to longer, unseen sensor traces. These findings suggest that state-space models could be a strong backbone for CPS-IoT foundation models for long-span reasoning tasks.
Authors: Vaneet Aggarwal, Shweta Jain, Subham Pokhriyal, Christopher John Quinn
Abstract: In this paper, we study bi-criteria optimization for combinatorial multi-armed bandits (CMAB) with bandit feedback. We propose a general framework that transforms discrete bi-criteria offline approximation algorithms into online algorithms with sublinear regret and cumulative constraint violation (CCV) guarantees. Our framework requires the offline algorithm to provide an $(\alpha, \beta)$-bi-criteria approximation ratio with $\delta$-resilience and utilize $\texttt{N}$ oracle calls to evaluate the objective and constraint functions. We prove that the proposed framework achieves sub-linear regret and CCV, with both bounds scaling as ${O}\left(\delta^{2/3} \texttt{N}^{1/3}T^{2/3}\log^{1/3}(T)\right)$. Crucially, the framework treats the offline algorithm with $\delta$-resilience as a black box, enabling flexible integration of existing approximation algorithms into the CMAB setting. To demonstrate its versatility, we apply our framework to several combinatorial problems, including submodular cover, submodular cost covering, and fair submodular maximization. These applications highlight the framework's broad utility in adapting offline guarantees to online bi-criteria optimization under bandit feedback.
Authors: Jerry Liu, Jessica Grogan, Owen Dugan, Ashish Rao, Simran Arora, Atri Rudra, Christopher R\'e
Abstract: This paper investigates whether sequence models can learn to perform numerical algorithms, e.g. gradient descent, on the fundamental problem of least squares. Our goal is to inherit two properties of standard algorithms from numerical analysis: (1) machine precision, i.e. we want to obtain solutions that are accurate to near floating point error, and (2) numerical generality, i.e. we want them to apply broadly across problem instances. We find that prior approaches using Transformers fail to meet these criteria, and identify limitations present in existing architectures and training procedures. First, we show that softmax Transformers struggle to perform high-precision multiplications, which prevents them from precisely learning numerical algorithms. Second, we identify an alternate class of architectures, comprised entirely of polynomials, that can efficiently represent high-precision gradient descent iterates. Finally, we investigate precision bottlenecks during training and address them via a high-precision training recipe that reduces stochastic gradient noise. Our recipe enables us to train two polynomial architectures, gated convolutions and linear attention, to perform gradient descent iterates on least squares problems. For the first time, we demonstrate the ability to train to near machine precision. Applied iteratively, our models obtain 100,000x lower MSE than standard Transformers trained end-to-end and they incur a 10,000x smaller generalization gap on out-of-distribution problems. We make progress towards end-to-end learning of numerical algorithms for least squares.
Authors: Amirabbas Afzali, Amirhossein Afsharrad, Seyed Shahabeddin Mousavi, Sanjay Lall
Abstract: Large Language Models (LLMs) have made significant strides in generating human-like responses, largely due to preference alignment techniques. However, these methods often assume unbiased human feedback, which is rarely the case in real-world scenarios. This paper introduces Content-Aware Noise-Resilient Preference Optimization (CNRPO), a novel framework that addresses multiple sources of content-dependent noise in preference learning. CNRPO employs a multi-objective optimization approach to separate true preferences from content-aware noises, effectively mitigating their impact. We leverage backdoor attack mechanisms to efficiently learn and control various noise sources within a single model. Theoretical analysis and extensive experiments on different synthetic noisy datasets demonstrate that CNRPO significantly improves alignment with primary human preferences while controlling for secondary noises and biases, such as response length and harmfulness.
Authors: Yuzheng Hu, Fan Wu, Ruicheng Xian, Yuhang Liu, Lydia Zakynthinou, Pritish Kamath, Chiyuan Zhang, David Forsyth
Abstract: We propose the notion of empirical privacy variance and study it in the context of differentially private fine-tuning of language models. Specifically, we show that models calibrated to the same $(\varepsilon, \delta)$-DP guarantee using DP-SGD with different hyperparameter configurations can exhibit significant variations in empirical privacy, which we quantify through the lens of memorization. We investigate the generality of this phenomenon across multiple dimensions and discuss why it is surprising and relevant. Through regression analysis, we examine how individual and composite hyperparameters influence empirical privacy. The results reveal a no-free-lunch trade-off: existing practices of hyperparameter tuning in DP-SGD, which focus on optimizing utility under a fixed privacy budget, often come at the expense of empirical privacy. To address this, we propose refined heuristics for hyperparameter selection that explicitly account for empirical privacy, showing that they are both precise and practically useful. Finally, we take preliminary steps to understand empirical privacy variance. We propose two hypotheses, identify limitations in existing techniques like privacy auditing, and outline open questions for future research.
Authors: Zhe Wang, Yanjun Qi
Abstract: Gradient optimization-based adversarial attack methods automate the learning of adversarial triggers to generate jailbreak prompts or leak system prompts. In this work, we take a closer look at the optimization objective of adversarial trigger learning and propose ATLA: Adversarial Trigger Learning with Augmented objectives. ATLA improves the negative log-likelihood loss used by previous studies into a weighted loss formulation that encourages the learned adversarial triggers to optimize more towards response format tokens. This enables ATLA to learn an adversarial trigger from just one query-response pair and the learned trigger generalizes well to other similar queries. We further design a variation to augment trigger optimization with an auxiliary loss that suppresses evasive responses. We showcase how to use ATLA to learn adversarial suffixes jailbreaking LLMs and to extract hidden system prompts. Empirically we demonstrate that ATLA consistently outperforms current state-of-the-art techniques, achieving nearly 100% success in attacking while requiring 80% fewer queries. ATLA learned jailbreak suffixes demonstrate high generalization to unseen queries and transfer well to new LLMs.
Authors: Wei-Wei Du, Yung-Chien Wang, Wen-Chih Peng
Abstract: The demand for property valuation has attracted significant attention from sellers, buyers, and customers applying for loans. Reviews of existing approaches have revealed shortcomings in terms of not being able to handle missing value situations, as well as lacking interpretability, which means they cannot be used in real-world applications. To address these challenges, we propose an LLM-Generated EXplainable PRopErty valuation SyStem with neighbor imputation called EXPRESS, which provides the customizable missing value imputation technique, and addresses the opaqueness of prediction by providing the feature-wise explanation generated by LLM. The dynamic nearest neighbor search finds similar properties depending on different application scenarios by property configuration set by users (e.g., house age as criteria for the house in rural areas, and locations for buildings in urban areas). Motivated by the human appraisal procedure, we generate feature-wise explanations to provide users with a more intuitive understanding of the prediction results.
Authors: Farhad Pourkamali-Anaraki
Abstract: Traditional neural network regression models provide only point estimates, failing to capture predictive uncertainty. Probabilistic neural networks (PNNs) address this limitation by producing output distributions, enabling the construction of prediction intervals. However, the common assumption of Gaussian output distributions often results in overly wide intervals, particularly in the presence of outliers or deviations from normality. To enhance the adaptability of PNNs, we propose t-Distributed Neural Networks (TDistNNs), which generate t-distributed outputs, parameterized by location, scale, and degrees of freedom. The degrees of freedom parameter allows TDistNNs to model heavy-tailed predictive distributions, improving robustness to non-Gaussian data and enabling more adaptive uncertainty quantification. We develop a novel loss function tailored for the t-distribution and derive efficient gradient computations for seamless integration into deep learning frameworks. Empirical evaluations on synthetic and real-world data demonstrate that TDistNNs improve the balance between coverage and interval width. Notably, for identical architectures, TDistNNs consistently produce narrower prediction intervals than Gaussian-based PNNs while maintaining proper coverage. This work contributes a flexible framework for uncertainty estimation in neural networks tasked with regression, particularly suited to settings involving complex output distributions.
Authors: Xiangfei Fang, Boying Wang, Chengying Huan, Shaonan Ma, Heng Zhang, Chen Zhao
Abstract: Hypergraph representation learning has garnered increasing attention across various domains due to its capability to model high-order relationships. Traditional methods often rely on hypergraph neural networks (HNNs) employing message passing mechanisms to aggregate vertex and hyperedge features. However, these methods are constrained by their dependence on hypergraph topology, leading to the challenge of imbalanced information aggregation, where high-degree vertices tend to aggregate redundant features, while low-degree vertices often struggle to capture sufficient structural features. To overcome the above challenges, we introduce HyperKAN, a novel framework for hypergraph representation learning that transcends the limitations of message-passing techniques. HyperKAN begins by encoding features for each vertex and then leverages Kolmogorov-Arnold Networks (KANs) to capture complex nonlinear relationships. By adjusting structural features based on similarity, our approach generates refined vertex representations that effectively addresses the challenge of imbalanced information aggregation. Experiments conducted on the real-world datasets demonstrate that HyperKAN significantly outperforms state of-the-art HNN methods, achieving nearly a 9% performance improvement on the Senate dataset.
Authors: Suchanuch Piriyasatit, Chaohao Yuan, Ercan Engin Kuruoglu
Abstract: Autism Spectrum Disorder (ASD) is a complex neurological condition characterized by varied developmental impairments, especially in communication and social interaction. Accurate and early diagnosis of ASD is crucial for effective intervention, which is enhanced by richer representations of brain activity. The brain functional connectome, which refers to the statistical relationships between different brain regions measured through neuroimaging, provides crucial insights into brain function. Traditional static methods often fail to capture the dynamic nature of brain activity, in contrast, dynamic brain connectome analysis provides a more comprehensive view by capturing the temporal variations in the brain. We propose BrainTWT, a novel dynamic network embedding approach that captures temporal evolution of the brain connectivity over time and considers also the dynamics between different temporal network snapshots. BrainTWT employs temporal random walks to capture dynamics across different temporal network snapshots and leverages the Transformer's ability to model long term dependencies in sequential data to learn the discriminative embeddings from these temporal sequences using temporal structure prediction tasks. The experimental evaluation, utilizing the Autism Brain Imaging Data Exchange (ABIDE) dataset, demonstrates that BrainTWT outperforms baseline methods in ASD classification.
Authors: Rui Xu, Dawen Yao, Yuzhuang Pian, Ruhui Cao, Yixin Fu, Xinru Yang, Ting Gan, Yonghong Liu
Abstract: Constructing high resolution air pollution maps at lower cost is crucial for sustainable city management and public health risk assessment. However, traditional fixed-site monitoring lacks spatial coverage, while mobile low-cost sensors exhibit significant data instability. This study integrates PM2.5 data from 320 taxi-mounted mobile low-cost sensors and 52 fixed monitoring stations to address these limitations. By employing the machine learning methods, an appropriate mapping relationship was established between fixed and mobile monitoring concentration. The resulting pollution maps achieved 500-meter spatial and 5-minute temporal resolutions, showing close alignment with fixed monitoring data (+4.35% bias) but significant deviation from raw mobile data (-31.77%). The fused map exhibits the fine-scale spatial variability also observed in the mobile pollution map, while showing the stable temporal variability closer to that of the fixed pollution map (fixed: 1.12 plus or minus 0.73%, mobile: 3.15 plus or minus 2.44%, mapped: 1.01 plus or minus 0.65%). These findings demonstrate the potential of large-scale mobile low-cost sensor networks for high-resolution air quality mapping, supporting targeted urban environmental governance and health risk mitigation.
Authors: Jonas Chris Ferrao, Dickson Dias, Sweta Morajkar, Manisha Gokuldas Fal Dessai
Abstract: Identifying transcription factor binding sites (TFBS) is crucial for understanding gene regulation, as these sites enable transcription factors (TFs) to bind to DNA and modulate gene expression. Despite advances in high-throughput sequencing, accurately identifying TFBS remains challenging due to the vast genomic data and complex binding patterns. GCBLANE, a graph-enhanced convolutional bidirectional Long Short-Term Memory (LSTM) attention network, is introduced to address this issue. It integrates convolutional, multi-head attention, and recurrent layers with a graph neural network to detect key features for TFBS prediction. On 690 ENCODE ChIP-Seq datasets, GCBLANE achieved an average AUC of 0.943, and on 165 ENCODE datasets, it reached an AUC of 0.9495, outperforming advanced models that utilize multimodal approaches, including DNA shape information. This result underscores GCBLANE's effectiveness compared to other methods. By combining graph-based learning with sequence analysis, GCBLANE significantly advances TFBS prediction.
Authors: Bocheng Wang, Chusheng Zeng, Mulin Chen, Xuelong Li
Abstract: Deep multi-view clustering incorporating graph learning has presented tremendous potential. Most methods encounter costly square time consumption w.r.t. data size. Theoretically, anchor-based graph learning can alleviate this limitation, but related deep models mainly rely on manual discretization approaches to select anchors, which indicates that 1) the anchors are fixed during model training and 2) they may deviate from the true cluster distribution. Consequently, the unreliable anchors may corrupt clustering results. In this paper, we propose the Deep Multi-view Anchor Clustering (DMAC) model that performs clustering in linear time. Concretely, the initial anchors are intervened by the positive-incentive noise sampled from Gaussian distribution, such that they can be optimized with a newly designed anchor learning loss, which promotes a clear relationship between samples and anchors. Afterwards, anchor graph convolution is devised to model the cluster structure formed by the anchors, and the mutual information maximization loss is built to provide cross-view clustering guidance. In this way, the learned anchors can better represent clusters. With the optimal anchors, the full sample graph is calculated to derive a discriminative embedding for clustering. Extensive experiments on several datasets demonstrate the superior performance and efficiency of DMAC compared to state-of-the-art competitors.
Authors: Sean Xiao, Sangwoo Park, Stefan Vlaski
Abstract: Stochastic first-order methods for empirical risk minimization employ gradient approximations based on sampled data in lieu of exact gradients. Such constructions introduce noise into the learning dynamics, which can be corrected through variance-reduction techniques. There is increasing evidence in the literature that in many modern learning applications noise can have a beneficial effect on optimization and generalization. To this end, the recently proposed variance-reduction technique, alpha-SVRG [Yin et al., 2023] allows for fine-grained control of the level of residual noise in the learning dynamics, and has been reported to empirically outperform both SGD and SVRG in modern deep learning scenarios. By focusing on strongly convex environments, we first provide a unified convergence rate expression for alpha-SVRG under fixed learning rate, which reduces to that of either SGD or SVRG by setting alpha=0 or alpha=1, respectively. We show that alpha-SVRG has faster convergence rate compared to SGD and SVRG under suitable choice of alpha. Simulation results on linear regression validate our theory.
Authors: Zhiyu Liang, Dongrui Cai, Chenyuan Zhang, Zheng Liang, Chen Liang, Bo Zheng, Shi Qiu, Jin Wang, Hongzhi Wang
Abstract: Model selection has been raised as an essential problem in the area of time series anomaly detection (TSAD), because there is no single best TSAD model for the highly heterogeneous time series in real-world applications. However, despite the success of existing model selection solutions that train a classification model (especially neural network, NN) using historical data as a selector to predict the correct TSAD model for each series, the NN-based selector learning methods used by existing solutions do not make full use of the knowledge in the historical data and require iterating over all training samples, which limits the accuracy and training speed of the selector. To address these limitations, we propose KDSelector, a novel knowledge-enhanced and data-efficient framework for learning the NN-based TSAD model selector, of which three key components are specifically designed to integrate available knowledge into the selector and dynamically prune less important and redundant samples during the learning. We develop a TSAD model selection system with KDSelector as the internal, to demonstrate how users improve the accuracy and training speed of their selectors by using KDSelector as a plug-and-play module. Our demonstration video is hosted at https://youtu.be/2uqupDWvTF0.
Authors: Patryk Marsza{\l}ek, Ulvi Movsum-zada, Oleksii Furman, Kamil Ksi\k{a}\.zek, Przemys{\l}aw Spurek, Marek \'Smieja
Abstract: In recent years, there has been a growing interest in explainable AI methods. We want not only to make accurate predictions using sophisticated neural networks but also to understand what the model's decision is based on. One of the fundamental levels of interpretability is to provide counterfactual examples explaining the rationale behind the decision and identifying which features, and to what extent, must be modified to alter the model's outcome. To address these requirements, we introduce HyConEx, a classification model based on deep hypernetworks specifically designed for tabular data. Owing to its unique architecture, HyConEx not only provides class predictions but also delivers local interpretations for individual data samples in the form of counterfactual examples that steer a given sample toward an alternative class. While many explainable methods generated counterfactuals for external models, there have been no interpretable classifiers simultaneously producing counterfactual samples so far. HyConEx achieves competitive performance on several metrics assessing classification accuracy and fulfilling the criteria of a proper counterfactual attack. This makes HyConEx a distinctive deep learning model, which combines predictions and explainers as an all-in-one neural network. The code is available at https://github.com/gmum/HyConEx.
Authors: Huajie Liang, Di Wang, Yuchao Lu, Mengke Song, Lei Liu, Ling An, Ying Liang, Xingjie Ma, Zhenyu Zhang, Chichun Zhou
Abstract: With the advancement of Industry 4.0, intelligent manufacturing extensively employs sensors for real-time multidimensional data collection, playing a crucial role in equipment monitoring, process optimisation, and efficiency enhancement. Industrial data exhibit characteristics such as multi-source heterogeneity, nonlinearity, strong coupling, and temporal interactions, while also being affected by noise interference. These complexities make it challenging for traditional anomaly detection methods to extract key features, impacting detection accuracy and stability. Traditional machine learning approaches often struggle with such complex data due to limitations in processing capacity and generalisation ability, making them inadequate for practical applications. While deep learning feature extraction modules have demonstrated remarkable performance in image and text processing, they remain ineffective when applied to multi-source heterogeneous industrial data lacking explicit correlations. Moreover, existing multi-source heterogeneous data processing techniques still rely on dimensionality reduction and feature selection, which can lead to information loss and difficulty in capturing high-order interactions. To address these challenges, this study applies the EAPCR and Time-EAPCR models proposed in previous research and introduces a new model, Time-EAPCR-T, where Transformer replaces the LSTM module in the time-series processing component of Time-EAPCR. This modification effectively addresses multi-source data heterogeneity, facilitates efficient multi-source feature fusion, and enhances the temporal feature extraction capabilities of multi-source industrial data.Experimental results demonstrate that the proposed method outperforms existing approaches across four industrial datasets, highlighting its broad application potential.
Authors: Lin-Chun Huang, Ching Chieh Tsao, Fang-Yi Su, Jung-Hsien Chiang
Abstract: Image generative models, particularly diffusion-based models, have surged in popularity due to their remarkable ability to synthesize highly realistic images. However, since these models are data-driven, they inherit biases from the training datasets, frequently leading to disproportionate group representations that exacerbate societal inequities. Traditionally, efforts to debiase these models have relied on predefined sensitive attributes, classifiers trained on such attributes, or large language models to steer outputs toward fairness. However, these approaches face notable drawbacks: predefined attributes do not adequately capture complex and continuous variations among groups. To address these issues, we introduce the Debiasing Diffusion Model (DDM), which leverages an indicator to learn latent representations during training, promoting fairness through balanced representations without requiring predefined sensitive attributes. This approach not only demonstrates its effectiveness in scenarios previously addressed by conventional techniques but also enhances fairness without relying on predefined sensitive attributes as conditions. In this paper, we discuss the limitations of prior bias mitigation techniques in diffusion-based models, elaborate on the architecture of the DDM, and validate the effectiveness of our approach through experiments.
Authors: Yancheng Wang, Changyu Liu, Yingzhen Yang
Abstract: Graph diffusion models have recently been proposed to synthesize entire graphs, such as molecule graphs. Although existing methods have shown great performance in generating entire graphs for graph-level learning tasks, no graph diffusion models have been developed to generate synthetic graph structures, that is, synthetic nodes and associated edges within a given graph, for node-level learning tasks. Inspired by the research in the computer vision literature using synthetic data for enhanced performance, we propose Diffusion on Graph (DoG), which generates synthetic graph structures to boost the performance of GNNs. The synthetic graph structures generated by DoG are combined with the original graph to form an augmented graph for the training of node-level learning tasks, such as node classification and graph contrastive learning (GCL). To improve the efficiency of the generation process, a Bi-Level Neighbor Map Decoder (BLND) is introduced in DoG. To mitigate the adverse effect of the noise introduced by the synthetic graph structures, a low-rank regularization method is proposed for the training of graph neural networks (GNNs) on the augmented graphs. Extensive experiments on various graph datasets for semi-supervised node classification and graph contrastive learning have been conducted to demonstrate the effectiveness of DoG with low-rank regularization. The code of DoG is available at https://github.com/Statistical-Deep-Learning/DoG.
Authors: Harshit
Abstract: Large Language Model (LLM) development has become increasingly centralized, limiting participation to well-resourced organizations. This paper introduces MoECollab, a novel framework leveraging Mixture of Experts (MoE) architecture to enable distributed, collaborative LLM development. By decomposing monolithic models into specialized expert modules coordinated by a trainable gating network, our framework allows diverse contributors to participate regardless of computational resources. We provide a complete technical implementation with mathematical foundations for expert dynamics, gating mechanisms, and integration strategies. Experiments on multiple datasets demonstrate that our approach achieves accuracy improvements of 3-7% over baseline models while reducing computational requirements by 34%. Expert specialization yields significant domain-specific gains, with improvements from 51% to 88% F1 score in general classification and from 23% to 44% accuracy in news categorization. We formalize the routing entropy optimization problem and demonstrate how proper regularization techniques lead to 14% higher expert utilization rates. These results validate MoECollab as an effective approach for democratizing LLM development through architecturally-supported collaboration.
Authors: Tao Feng, Yihang Sun, Jiaxuan You
Abstract: The powerful capabilities of Large Language Models (LLMs) have led to their growing use in evaluating human-generated content, particularly in evaluating research ideas within academic settings. Existing solutions primarily rely on prompt-based LLM methods or fine-tuned lightweight language models for idea evaluation. However, these methods are often unstable and struggle to comprehend the complex semantic information embedded in the ideas, impeding their ability to perform high-quality evaluations. To address the above challenges, we propose GraphEval, a lightweight graph-based LLM framework for idea evaluation. Our insight is that a complex idea can be broken down into comprehensible viewpoint nodes using prompts from small LLMs. These viewpoint nodes can then be linked together through edges created from LLM-based relation extraction and/or BERT similarity scores. The created viewpoint-graph can be used to conveniently propagate scores across view-nodes to improve the robustness of the idea evaluations. In particular, we propose two lightweight graph-based methods for idea evaluation: (1) GraphEval-LP: a training-free label propagation algorithm that propagates evaluation scores from known view-nodes to unknown nodes; (2) GraphEval-GNN: a Graph Neural Networks (GNN) that is trained to predict the evaluation scores given the observed graph with minimal computation resources. Moreover, to overcome LLM's limitation in objectively assessing the novelty of ideas, we further propose a novelty detection model to GraphEval-GNN to enhance its capability in judging idea novelty. Experiments on two datasets show GraphEval improves F1 scores by at least 14% with low computation and API costs. Additionally, GraphEval can effectively detect plagiarized ideas.
Authors: Kunyang Sun, Dorian Bagni, Joseph M. Cavanagh, Yingze Wang, Jacob M. Sawyer, Andrew Gritsevskiy, Teresa Head-Gordon
Abstract: Generative machine learning models for small molecule drug discovery have shown immense promise, but many molecules generated by this approach are too difficult to synthesize to be worth further investigation or further development. We present a novel approach by fine-tuning Meta's Llama3 large language models (LLMs) to create SynLlama, which generates full synthetic pathways made of commonly accessible Enamine building blocks and robust organic reaction templates. SynLlama explores a large synthesizable space using significantly less data compared to other state-of-the-art methods, and offers strong performance in bottom-up synthesis, synthesizable analog generation, and hit expansion, offering medicinal chemists a valuable tool for drug discovery developments. We find that SynLlama can effectively generalize to unseen yet purchasable building blocks, meaning that its reconstruction capabilities extend to a broader synthesizable chemical space than the training data.
Authors: Khayrul Islam, Ryan F. Forelli, Jianzhong Han, Deven Bhadane, Jian Huang, Joshua C. Agar, Nhan Tran, Seda Ogrenci, Yaling Liu
Abstract: Precise cell classification is essential in biomedical diagnostics and therapeutic monitoring, particularly for identifying diverse cell types involved in various diseases. Traditional cell classification methods such as flow cytometry depend on molecular labeling which is often costly, time-intensive, and can alter cell integrity. To overcome these limitations, we present a label-free machine learning framework for cell classification, designed for real-time sorting applications using bright-field microscopy images. This approach leverages a teacher-student model architecture enhanced by knowledge distillation, achieving high efficiency and scalability across different cell types. Demonstrated through a use case of classifying lymphocyte subsets, our framework accurately classifies T4, T8, and B cell types with a dataset of 80,000 preprocessed images, accessible via an open-source Python package for easy adaptation. Our teacher model attained 98\% accuracy in differentiating T4 cells from B cells and 93\% accuracy in zero-shot classification between T8 and B cells. Remarkably, our student model operates with only 0.02\% of the teacher model's parameters, enabling field-programmable gate array (FPGA) deployment. Our FPGA-accelerated student model achieves an ultra-low inference latency of just 14.5~$\mu$s and a complete cell detection-to-sorting trigger time of 24.7~$\mu$s, delivering 12x and 40x improvements over the previous state-of-the-art real-time cell analysis algorithm in inference and total latency, respectively, while preserving accuracy comparable to the teacher model. This framework provides a scalable, cost-effective solution for lymphocyte classification, as well as a new SOTA real-time cell sorting implementation for rapid identification of subsets using in situ deep learning on off-the-shelf computing hardware.
Authors: Vrushank Ahire, Kunal Shah, Mudasir Nazir Khan, Nikhil Pakhale, Lownish Rai Sookha, M. A. Ganaie, Abhinav Dhall
Abstract: This paper introduces MAVEN (Multi-modal Attention for Valence-Arousal Emotion Network), a novel architecture for dynamic emotion recognition through dimensional modeling of affect. The model uniquely integrates visual, audio, and textual modalities via a bi-directional cross-modal attention mechanism with six distinct attention pathways, enabling comprehensive interactions between all modality pairs. Our proposed approach employs modality-specific encoders to extract rich feature representations from synchronized video frames, audio segments, and transcripts. The architecture's novelty lies in its cross-modal enhancement strategy, where each modality representation is refined through weighted attention from other modalities, followed by self-attention refinement through modality-specific encoders. Rather than directly predicting valence-arousal values, MAVEN predicts emotions in a polar coordinate form, aligning with psychological models of the emotion circumplex. Experimental evaluation on the Aff-Wild2 dataset demonstrates the effectiveness of our approach, with performance measured using Concordance Correlation Coefficient (CCC). The multi-stage architecture demonstrates superior ability to capture the complex, nuanced nature of emotional expressions in conversational videos, advancing the state-of-the-art (SOTA) in continuous emotion recognition in-the-wild. Code can be found at: https://github.com/Vrushank-Ahire/MAVEN_8th_ABAW.
Authors: Amin Banayeeanzade, Mohammad Rostami
Abstract: Continual learning is crucial for creating AI agents that can learn and improve themselves autonomously. A primary challenge in continual learning is to learn new tasks without losing previously learned knowledge. Current continual learning methods primarily focus on enabling a neural network with mechanisms that mitigate forgetting effects. Inspired by the two distinct systems in the human brain, System 1 and System 2, we propose a Neuro-Symbolic Brain-Inspired Continual Learning (NeSyBiCL) framework that incorporates two subsystems to solve continual learning: A neural network model responsible for quickly adapting to the most recent task, together with a symbolic reasoner responsible for retaining previously acquired knowledge from previous tasks. Moreover, we design an integration mechanism between these components to facilitate knowledge transfer from the symbolic reasoner to the neural network. We also introduce two compositional continual learning benchmarks and demonstrate that NeSyBiCL is effective and leads to superior performance compared to continual learning methods that merely rely on neural architectures to address forgetting.
Authors: Dmitry Kovalev
Abstract: Optimization with matrix gradient orthogonalization has recently demonstrated impressive results in the training of deep neural networks (Jordan et al., 2024; Liu et al., 2025). In this paper, we provide a theoretical analysis of this approach. In particular, we show that the orthogonalized gradient method can be seen as a first-order trust-region optimization method, where the trust-region is defined in terms of the matrix spectral norm. Motivated by this observation, we provide the first theoretical analysis of the stochastic non-Euclidean trust-region gradient method with momentum, which recovers the Muon optimizer (Jordan et al., 2024) as a special case. In addition, we establish the convergence of the normalized SGD with momentum (Cutkosky and Mehta, 2020) in the constrained and composite setting, show that its iteration complexity of finding an $\varepsilon$-accurate solution can be improved from $\mathcal{O}(\varepsilon^{-3.5})$ to $\mathcal{O}(\varepsilon^{-3})$ under the star-convexity assumption, and obtain similar results for the Muon algorithm. Finally, our theoretical findings provide an explanation for the practical superiority of Muon compared to the Orthogonal-SGDM algorithm of Tuddenham et al. (2022).
Authors: Andreas Teller, Uta Pigorsch, Christian Pigorsch
Abstract: Forecasting the volatility of financial assets is essential for various financial applications. This paper addresses the challenging task of forecasting the volatility of financial assets with limited historical data, such as new issues or spin-offs, by proposing a multi-source transfer learning approach. Specifically, we exploit complementary source data of assets with a substantial historical data record by selecting source time series instances that are most similar to the limited target data of the new issue/spin-off. Based on these instances and the target data, we estimate linear and non-linear realized volatility models and compare their forecasting performance to forecasts of models trained exclusively on the target data, and models trained on the entire source and target data. The results show that our transfer learning approach outperforms the alternative models and that the integration of complementary data is also beneficial immediately after the initial trading day of the new issue/spin-off.
Authors: Hao Mark Chen, Shell Xu Hu, Wayne Luk, Timothy Hospedales, Hongxiang Fan
Abstract: Model merging has emerged as a promising approach for multi-task learning (MTL), offering a data-efficient alternative to conventional fine-tuning. However, with the rapid development of the open-source AI ecosystem and the increasing availability of fine-tuned foundation models, existing model merging methods face two key limitations: (i) They are primarily designed for in-house fine-tuned models, making them less adaptable to diverse model sources with partially unknown model and task information, (ii) They struggle to scale effectively when merging numerous model checkpoints. To address these challenges, we formulate model merging as a constrained optimization problem and introduce a novel approach: Frank-Wolfe Merging (FW-Merging). Inspired by Frank-Wolfe optimization, our approach iteratively selects the most relevant model in the pool to minimize a linear approximation of the objective function and then executes a local merging similar to the Frank-Wolfe update. The objective function is designed to capture the desired behavior of the target-merged model, while the fine-tuned candidate models define the constraint set. More importantly, FW-Merging serves as an orthogonal technique for existing merging methods, seamlessly integrating with them to further enhance accuracy performance. Our experiments show that FW-Merging scales across diverse model sources, remaining stable with 16 irrelevant models and improving by 15.3% with 16 relevant models on 20 CV tasks, while maintaining constant memory overhead, unlike the linear overhead of data-informed merging methods. Compared with the state-of-the-art approaches, FW-Merging surpasses the data-free merging method by 32.8% and outperforms the data-informed Adamerging by 8.39% when merging 20 ViT models.
Authors: Arthur Corr\^ea, Crist\'ov\~ao Silva, Liming Xu, Alexandra Brintrup, Samuel Moniz
Abstract: This paper introduces TuneNSearch, a hybrid transfer learning and local search approach for addressing different variants of vehicle routing problems (VRP). Recently, multi-task learning has gained much attention for solving VRP variants. However, this adaptability often compromises the performance of the models. To address this challenge, we first pre-train a reinforcement learning model on the multi-depot VRP, followed by a short fine-tuning phase to adapt it to different variants. By leveraging the complexity of the multi-depot VRP, the pre-trained model learns richer node representations and gains more transferable knowledge compared to models trained on simpler routing problems, such as the traveling salesman problem. TuneNSearch employs, in the first stage, a Transformer-based architecture, augmented with a residual edge-graph attention network to capture the impact of edge distances and residual connections between layers. This architecture allows for a more precise capture of graph-structured data, improving the encoding of VRP's features. After inference, our model is also coupled with a second stage composed of a local search algorithm, which yields substantial performance gains with minimal computational overhead added. Results show that TuneNSearch outperforms many existing state-of-the-art models trained for each VRP variant, requiring only one-fifth of the training epochs. Our approach demonstrates strong generalization, achieving high performance across different tasks, distributions and problem sizes, thus addressing a long-standing gap in the literature.
Authors: Liangyu Wang, Jie Ren, Hang Xu, Junxiao Wang, Huanyi Xie, David E. Keyes, Di Wang
Abstract: Fine-tuning large pre-trained LLMs generally demands extensive GPU memory. Traditional first-order optimizers like SGD encounter substantial difficulties due to increased memory requirements from storing activations and gradients during both the forward and backward phases as the model size expands. Alternatively, zeroth-order (ZO) techniques can compute gradients using just forward operations, eliminating the need to store activations. Furthermore, by leveraging CPU capabilities, it's feasible to enhance both the memory and processing power available to a single GPU. We propose a novel framework, ZO2 (Zeroth-Order Offloading), for efficient zeroth-order fine-tuning of LLMs with only limited GPU memory. Our framework dynamically shifts model parameters between the CPU and GPU as required, optimizing computation flow and maximizing GPU usage by minimizing downtime. This integration of parameter adjustments with ZO's double forward operations reduces unnecessary data movement, enhancing the fine-tuning efficacy. Additionally, our framework supports an innovative low-bit precision approach in AMP mode to streamline data exchanges between the CPU and GPU. Employing this approach allows us to fine-tune extraordinarily large models, such as the OPT-175B with more than 175 billion parameters, on a mere 18GB GPU--achievements beyond the reach of traditional methods. Moreover, our framework achieves these results with almost no additional time overhead and absolutely no accuracy loss compared to standard zeroth-order methods. ZO2's code has been open-sourced in https://github.com/liangyuwang/zo2.
Authors: Roozbeh Siyadatzadeh, Mohsen Ansari, Muhammad Shafique, Alireza Ejlali
Abstract: Embedded systems power many modern applications and must often meet strict reliability, real-time, thermal, and power requirements. Task replication can improve reliability by duplicating a task's execution to handle transient and permanent faults, but blindly applying replication often leads to excessive overhead and higher temperatures. Existing design-time methods typically choose the number of replicas based on worst-case conditions, which can waste resources under normal operation. In this paper, we present RL-TIME, a reinforcement learning-based approach that dynamically decides the number of replicas according to actual system conditions. By considering both the reliability target and a core-level Thermal Safe Power (TSP) constraint at run-time, RL-TIME adapts the replication strategy to avoid unnecessary overhead and overheating. Experimental results show that, compared to state-of-the-art methods, RL-TIME reduces power consumption by 63%, increases schedulability by 53%, and respects TSP 72% more often.
Authors: Lachlan Simpson, Federico Costanza, Kyle Millar, Adriel Cheng, Cheng-Chew Lim, Hong Gunn Chew
Abstract: Classical adversarial attacks are phrased as a constrained optimisation problem. Despite the efficacy of a constrained optimisation approach to adversarial attacks, one cannot trace how an adversarial point was generated. In this work, we propose an algebraic approach to adversarial attacks and study the conditions under which one can generate adversarial examples for post-hoc explainability models. Phrasing neural networks in the framework of geometric deep learning, algebraic adversarial attacks are constructed through analysis of the symmetry groups of neural networks. Algebraic adversarial examples provide a mathematically tractable approach to adversarial examples. We validate our approach of algebraic adversarial examples on two well-known and one real-world dataset.
Authors: Jacqueline L. Mitchell, Brian Hyeongseok Kim, Chenyu Zhou, Chao Wang
Abstract: LLMs have demonstrated impressive capabilities in code generation and comprehension, but their potential in being able to perform program analysis in a formal, automatic manner remains under-explored. To that end, we systematically investigate whether LLMs can reason about programs using a program analysis framework called abstract interpretation. We prompt LLMs to follow two different strategies, denoted as Compositional and Fixed Point Equation, to formally reason in the style of abstract interpretation, which has never been done before to the best of our knowledge. We validate our approach using state-of-the-art LLMs on 22 challenging benchmark programs from the Software Verification Competition (SV-COMP) 2019 dataset, widely used in program analysis. Our results show that our strategies are able to elicit abstract interpretation-based reasoning in the tested models, but LLMs are susceptible to logical errors, especially while interpreting complex program structures, as well as general hallucinations. This highlights key areas for improvement in the formal reasoning capabilities of LLMs.
Authors: Philip Quirke, Clement Neo, Abir Harrasse, Dhruv Nathawani, Amir Abdullah
Abstract: Mechanistic interpretability research faces a gap between analyzing simple circuits in toy tasks and discovering features in large models. To bridge this gap, we propose text-to-SQL generation as an ideal task to study, as it combines the formal structure of toy tasks with real-world complexity. We introduce TinySQL, a synthetic dataset progressing from basic to advanced SQL operations, and train models ranging from 33M to 1B parameters to establish a comprehensive testbed for interpretability. We apply multiple complementary interpretability techniques, including edge attribution patching and sparse autoencoders, to identify minimal circuits and components supporting SQL generation. Our analysis reveals both the potential and limitations of current interpretability methods, showing how circuits can vary even across similar queries. Lastly, we demonstrate how mechanistic interpretability can identify flawed heuristics in models and improve synthetic dataset design. Our work provides a comprehensive framework for evaluating and advancing interpretability techniques while establishing clear boundaries for their reliable application.
Authors: Patrick Hytla, ran T. A. Nghia, Duy Nhat Phan, Andrew Rice
Abstract: Matrix completion is fundamental for predicting missing data with a wide range of applications in personalized healthcare, e-commerce, recommendation systems, and social network analysis. Traditional matrix completion approaches typically assume centralized data storage, which raises challenges in terms of computational efficiency, scalability, and user privacy. In this paper, we address the problem of federated matrix completion, focusing on scenarios where user-specific data is distributed across multiple clients, and privacy constraints are uncompromising. Federated learning provides a promising framework to address these challenges by enabling collaborative learning across distributed datasets without sharing raw data. We propose \texttt{FedMC-ADMM} for solving federated matrix completion problems, a novel algorithmic framework that combines the Alternating Direction Method of Multipliers with a randomized block-coordinate strategy and alternating proximal gradient steps. Unlike existing federated approaches, \texttt{FedMC-ADMM} effectively handles multi-block nonconvex and nonsmooth optimization problems, allowing efficient computation while preserving user privacy. We analyze the theoretical properties of our algorithm, demonstrating subsequential convergence and establishing a convergence rate of $\mathcal{O}(K^{-1/2})$, leading to a communication complexity of $\mathcal{O}(\epsilon^{-2})$ for reaching an $\epsilon$-stationary point. This work is the first to establish these theoretical guarantees for federated matrix completion in the presence of multi-block variables. To validate our approach, we conduct extensive experiments on real-world datasets, including MovieLens 1M, 10M, and Netflix. The results demonstrate that \texttt{FedMC-ADMM} outperforms existing methods in terms of convergence speed and testing accuracy.
Authors: Jianliang He, Xintian Pan, Siyu Chen, Zhuoran Yang
Abstract: We study how multi-head softmax attention models are trained to perform in-context learning on linear data. Through extensive empirical experiments and rigorous theoretical analysis, we demystify the emergence of elegant attention patterns: a diagonal and homogeneous pattern in the key-query (KQ) weights, and a last-entry-only and zero-sum pattern in the output-value (OV) weights. Remarkably, these patterns consistently appear from gradient-based training starting from random initialization. Our analysis reveals that such emergent structures enable multi-head attention to approximately implement a debiased gradient descent predictor -- one that outperforms single-head attention and nearly achieves Bayesian optimality up to proportional factor. Furthermore, compared to linear transformers, the softmax attention readily generalizes to sequences longer than those seen during training. We also extend our study to scenarios with non-isotropic covariates and multi-task linear regression. In the former, multi-head attention learns to implement a form of pre-conditioned gradient descent. In the latter, we uncover an intriguing regime where the interplay between head number and task number triggers a superposition phenomenon that efficiently resolves multi-task in-context learning. Our results reveal that in-context learning ability emerges from the trained transformer as an aggregated effect of its architecture and the underlying data distribution, paving the way for deeper understanding and broader applications of in-context learning.
Authors: Yu Xia, Zhiqiang Xu
Abstract: This paper investigates the ability of finite samples to identify two-layer irreducible shallow networks with various nonlinear activation functions, including rectified linear units (ReLU) and analytic functions such as the logistic sigmoid and hyperbolic tangent. An ``irreducible" network is one whose function cannot be represented by another network with fewer neurons. For ReLU activation functions, we first establish necessary and sufficient conditions for determining the irreducibility of a network. Subsequently, we prove a negative result: finite samples are insufficient for definitive identification of any irreducible ReLU shallow network. Nevertheless, we demonstrate that for a given irreducible network, one can construct a finite set of sampling points that can distinguish it from other network with the same neuron count. Conversely, for logistic sigmoid and hyperbolic tangent activation functions, we provide a positive result. We construct finite samples that enable the recovery of two-layer irreducible shallow analytic networks. To the best of our knowledge, this is the first study to investigate the exact identification of two-layer irreducible networks using finite sample function values. Our findings provide insights into the comparative performance of networks with different activation functions under limited sampling conditions.
Authors: Longfei Wei, Fang Sheng, Jianfei Zhang
Abstract: Artificial intelligence (AI) has significantly improved medical screening accuracy, particularly in cancer detection and risk assessment. However, traditional classification metrics often fail to account for imbalanced data, varying performance across cohorts, and patient-level inconsistencies, leading to biased evaluations. We propose the Cohort-Attention Evaluation Metrics (CAT) framework to address these challenges. CAT introduces patient-level assessment, entropy-based distribution weighting, and cohort-weighted sensitivity and specificity. Key metrics like CATSensitivity (CATSen), CATSpecificity (CATSpe), and CATMean ensure balanced and fair evaluation across diverse populations. This approach enhances predictive reliability, fairness, and interpretability, providing a robust evaluation method for AI-driven medical screening models.
Authors: Linjian Meng, Youzhi Zhang, Zhenxing Ge, Tianpei Yang, Yang Gao
Abstract: Counterfactual Regret Minimization (CFR) algorithms are widely used to compute a Nash equilibrium (NE) in two-player zero-sum imperfect-information extensive-form games (IIGs). Among them, Predictive CFR$^+$ (PCFR$^+$) is particularly powerful, achieving an exceptionally fast empirical convergence rate via the prediction in many games. However, the empirical convergence rate of PCFR$^+$ would significantly degrade if the prediction is inaccurate, leading to unstable performance on certain IIGs. To enhance the robustness of PCFR$^+$, we propose a novel variant, Asynchronous PCFR$^+$ (APCFR$^+$), which employs an adaptive asynchronization of step-sizes between the updates of implicit and explicit accumulated counterfactual regrets to mitigate the impact of the prediction inaccuracy on convergence. We present a theoretical analysis demonstrating why APCFR$^+$ can enhance the robustness. Finally, we propose a simplified version of APCFR$^+$ called Simple APCFR$^+$ (SAPCFR$^+$), which uses a fixed asynchronization of step-sizes to simplify the implementation that only needs a single-line modification of the original PCFR+. Interestingly, SAPCFR$^+$ achieves a constant-factor lower theoretical regret bound than PCFR$^+$ in the worst case. Experimental results demonstrate that (i) both APCFR$^+$ and SAPCFR$^+$ outperform PCFR$^+$ in most of the tested games, as well as (ii) SAPCFR$^+$ achieves a comparable empirical convergence rate with APCFR$^+$.
Authors: Yechao Zhang, Yingzhe Xu, Junyu Shi, Leo Yu Zhang, Shengshan Hu, Minghui Li, Yanjun Zhang
Abstract: Deep neural networks (DNNs) are susceptible to universal adversarial perturbations (UAPs). These perturbations are meticulously designed to fool the target model universally across all sample classes. Unlike instance-specific adversarial examples (AEs), generating UAPs is more complex because they must be generalized across a wide range of data samples and models. Our research reveals that existing universal attack methods, which optimize UAPs using DNNs with static model parameter snapshots, do not fully leverage the potential of DNNs to generate more effective UAPs. Rather than optimizing UAPs against static DNN models with a fixed training set, we suggest using dynamic model-data pairs to generate UAPs. In particular, we introduce a dynamic maximin optimization strategy, aiming to optimize the UAP across a variety of optimal model-data pairs. We term this approach DM-UAP. DM-UAP utilizes an iterative max-min-min optimization framework that refines the model-data pairs, coupled with a curriculum UAP learning algorithm to examine the combined space of model parameters and data thoroughly. Comprehensive experiments on the ImageNet dataset demonstrate that the proposed DM-UAP markedly enhances both cross-sample universality and cross-model transferability of UAPs. Using only 500 samples for UAP generation, DM-UAP outperforms the state-of-the-art approach with an average increase in fooling ratio of 12.108%.
Authors: Chen Li, Huidong Tang, Ye Zhu, Yoshihiro Yamanishi
Abstract: Generating molecules with desired chemical properties presents a critical challenge in fields such as chemical synthesis and drug discovery. Recent advancements in artificial intelligence (AI) and deep learning have significantly contributed to data-driven molecular generation. However, challenges persist due to the inherent sensitivity of simplified molecular input line entry system (SMILES) representations and the difficulties in applying generative adversarial networks (GANs) to discrete data. This study introduces RL-MolGAN, a novel Transformer-based discrete GAN framework designed to address these challenges. Unlike traditional Transformer architectures, RL-MolGAN utilizes a first-decoder-then-encoder structure, facilitating the generation of drug-like molecules from both $de~novo$ and scaffold-based designs. In addition, RL-MolGAN integrates reinforcement learning (RL) and Monte Carlo tree search (MCTS) techniques to enhance the stability of GAN training and optimize the chemical properties of the generated molecules. To further improve the model's performance, RL-MolWGAN, an extension of RL-MolGAN, incorporates Wasserstein distance and mini-batch discrimination, which together enhance the stability of the GAN. Experimental results on two widely used molecular datasets, QM9 and ZINC, validate the effectiveness of our models in generating high-quality molecular structures with diverse and desirable chemical properties.
Authors: Mohammad Wahiduzzaman Khan, Sheng Chen, Ilya Mironov, Leizhen Zhang, Rabib Noor
Abstract: Model memorization has implications for both the generalization capacity of machine learning models and the privacy of their training data. This paper investigates label memorization in binary classification models through two novel passive label inference attacks (BLIA). These attacks operate passively, relying solely on the outputs of pre-trained models, such as confidence scores and log-loss values, without interacting with or modifying the training process. By intentionally flipping 50% of the labels in controlled subsets, termed "canaries," we evaluate the extent of label memorization under two conditions: models trained without label differential privacy (Label-DP) and those trained with randomized response-based Label-DP. Despite the application of varying degrees of Label-DP, the proposed attacks consistently achieve success rates exceeding 50%, surpassing the baseline of random guessing and conclusively demonstrating that models memorize training labels, even when these labels are deliberately uncorrelated with the features.
Authors: Kairong Luo, Haodong Wen, Shengding Hu, Zhenbo Sun, Zhiyuan Liu, Maosong Sun, Kaifeng Lyu, Wenguang Chen
Abstract: Training large models is both resource-intensive and time-consuming, making it crucial to understand the quantitative relationship between model performance and hyperparameters. In this paper, we present an empirical law that describes how the pretraining loss of large language models evolves under different learning rate schedules, such as constant, cosine, and step decay schedules. Our proposed law takes a multi-power form, combining a power law based on the sum of learning rates and additional power laws to account for a loss reduction effect induced by learning rate decay. We extensively validate this law on various model sizes and architectures, and demonstrate that after fitting on a few learning rate schedules, the law accurately predicts the loss curves for unseen schedules of different shapes and horizons. Moreover, by minimizing the predicted final pretraining loss across learning rate schedules, we are able to find a schedule that outperforms the widely used cosine learning rate schedule. Interestingly, this automatically discovered schedule bears some resemblance to the recently proposed Warmup-Stable-Decay (WSD) schedule (Hu et al, 2024) but achieves a slightly lower final loss. We believe these results could offer valuable insights for understanding the dynamics of pretraining and designing learning rate schedules to improve efficiency.
Authors: Mehdi Makni, Kayhan Behdin, Gabriel Afriat, Zheng Xu, Sergei Vassilvitskii, Natalia Ponomareva, Hussein Hazimeh, Rahul Mazumder
Abstract: Differentially private stochastic gradient descent (DP-SGD) is broadly considered to be the gold standard for training and fine-tuning neural networks under differential privacy (DP). With the increasing availability of high-quality pre-trained model checkpoints (e.g., vision and language models), fine-tuning has become a popular strategy. However, despite recent progress in understanding and applying DP-SGD for private transfer learning tasks, significant challenges remain -- most notably, the performance gap between models fine-tuned with DP-SGD and their non-private counterparts. Sparse fine-tuning on private data has emerged as an alternative to full-model fine-tuning; recent work has shown that privately fine-tuning only a small subset of model weights and keeping the rest of the weights fixed can lead to better performance. In this work, we propose a new approach for sparse fine-tuning of neural networks under DP. Existing work on private sparse finetuning often used fixed choice of trainable weights (e.g., updating only the last layer), or relied on public model's weights to choose the subset of weights to modify. Such choice of weights remains suboptimal. In contrast, we explore an optimization-based approach, where our selection method makes use of the private gradient information, while using off the shelf privacy accounting techniques. Our numerical experiments on several computer vision models and datasets show that our selection method leads to better prediction accuracy, compared to full-model private fine-tuning or existing private sparse fine-tuning approaches.
Authors: Xian-Rong Zhang, Yue-Jiao Gong, Zhiguang Cao, Jun Zhang
Abstract: In recent years, there has been a growing interest in data-driven evolutionary algorithms (DDEAs) employing surrogate models to approximate the objective functions with limited data. However, current DDEAs are primarily designed for lower-dimensional problems and their performance drops significantly when applied to large-scale optimization problems (LSOPs). To address the challenge, this paper proposes an offline DDEA named DSKT-DDEA. DSKT-DDEA leverages multiple islands that utilize different data to establish diverse surrogate models, fostering diverse subpopulations and mitigating the risk of premature convergence. In the intra-island optimization phase, a semi-supervised learning method is devised to fine-tune the surrogates. It not only facilitates data argumentation, but also incorporates the distribution information gathered during the search process to align the surrogates with the evolving local landscapes. Then, in the inter-island knowledge transfer phase, the algorithm incorporates an adaptive strategy that periodically transfers individual information and evaluates the transfer effectiveness in the new environment, facilitating global optimization efficacy. Experimental results demonstrate that our algorithm is competitive with state-of-the-art DDEAs on problems with up to 1000 dimensions, while also exhibiting decent parallelism and scalability. Our DSKT-DDEA is open-source and accessible at: https://github.com/LabGong/DSKT-DDEA.
Authors: Maximilian Kirsch, Jakob Wernicke, Pawan Datta, Christine Preisach
Abstract: Climate change has increased the vulnerability of forests to insect-related damage, resulting in widespread forest loss in Central Europe and highlighting the need for effective, continuous monitoring systems. Remote sensing based forest health monitoring, oftentimes, relies on supervised machine learning algorithms that require labeled training data. Monitoring temporal patterns through time series analysis offers a potential alternative for earlier detection of disturbance but requires substantial storage resources. This study investigates the potential of a Deep Learning algorithm based on a Long Short Term Memory (LSTM) Autoencoder for the detection of anomalies in forest health (e.g. bark beetle outbreaks), utilizing Sentinel-2 time series data. This approach is an alternative to supervised machine learning methods, avoiding the necessity for labeled training data. Furthermore, it is more memory-efficient than other time series analysis approaches, as a robust model can be created using only a 26-week-long time series as input. In this study, we monitored pure stands of spruce in Thuringia, Germany, over a 7-year period from 2018 to the end of 2024. Our best model achieved a detection accuracy of 87% on test data and was able to detect 61% of all anomalies at a very early stage (more than a month before visible signs of forest degradation). Compared to another widely used time series break detection algorithm - BFAST (Breaks For Additive Season and Trend), our approach consistently detected higher percentage of anomalies at an earlier stage. These findings suggest that LSTM-based Autoencoders could provide a promising, resource-efficient approach to forest health monitoring, enabling more timely responses to emerging threats.
Authors: Haiyang Guo, Fanhu Zeng, Fei Zhu, Wenzhuo Liu, Da-Han Wang, Jian Xu, Xu-Yao Zhang, Cheng-Lin Liu
Abstract: A vast amount of instruction tuning data is crucial for the impressive performance of Large Multimodal Models (LMMs), but the associated computational costs and data collection demands during supervised fine-tuning make it impractical for most researchers. Federated learning (FL) has the potential to leverage all distributed data and training resources to reduce the overhead of joint training. However, most existing methods assume a fixed number of tasks, while in real-world scenarios, clients continuously encounter new knowledge and often struggle to retain old tasks due to memory constraints. In this work, we introduce the Federated Continual Instruction Tuning (FCIT) benchmark to model this real-world challenge. Our benchmark includes two realistic scenarios, encompassing four different settings and twelve carefully curated instruction tuning datasets. To address the challenges posed by FCIT, we propose dynamic knowledge organization to effectively integrate updates from different tasks during training and subspace selective activation to allocate task-specific output during inference. Extensive experimental results demonstrate that our proposed method significantly enhances model performance across varying levels of data heterogeneity and catastrophic forgetting. Our source code and dataset will be made publicly available.
Authors: Sabino Francesco Roselli, Eibe Frank
Abstract: Model trees provide an appealing way to perform interpretable machine learning for both classification and regression problems. In contrast to ``classic'' decision trees with constant values in their leaves, model trees can use linear combinations of predictor variables in their leaf nodes to form predictions, which can help achieve higher accuracy and smaller trees. Typical algorithms for learning model trees from training data work in a greedy fashion, growing the tree in a top-down manner by recursively splitting the data into smaller and smaller subsets. Crucially, the selected splits are only locally optimal, potentially rendering the tree overly complex and less accurate than a tree whose structure is globally optimal for the training data. In this paper, we empirically investigate the effect of constructing globally optimal model trees for classification and regression with linear support vector machines at the leaf nodes. To this end, we present mixed-integer linear programming formulations to learn optimal trees, compute such trees for a large collection of benchmark data sets, and compare their performance against greedily grown model trees in terms of interpretability and accuracy. We also compare to classic optimal and greedily grown decision trees, random forests, and support vector machines. Our results show that optimal model trees can achieve competitive accuracy with very small trees. We also investigate the effect on the accuracy of replacing axis-parallel splits with multivariate ones, foregoing interpretability while potentially obtaining greater accuracy.
Authors: Aref Einizade, Dorina Thanou, Fragkiskos D. Malliaros, Jhony H. Giraldo
Abstract: Simplicial complexes provide a powerful framework for modeling high-order interactions in structured data, making them particularly suitable for applications such as trajectory prediction and mesh processing. However, existing simplicial neural networks (SNNs), whether convolutional or attention-based, rely primarily on discrete filtering techniques, which can be restrictive. In contrast, partial differential equations (PDEs) on simplicial complexes offer a principled approach to capture continuous dynamics in such structures. In this work, we introduce COntinuous SiMplicial neural netwOrkS (COSMOS), a novel SNN architecture derived from PDEs on simplicial complexes. We provide theoretical and experimental justifications of COSMOS's stability under simplicial perturbations. Furthermore, we investigate the over-smoothing phenomenon, a common issue in geometric deep learning, demonstrating that COSMOS offers better control over this effect than discrete SNNs. Our experiments on real-world datasets of ocean trajectory prediction and regression on partial deformable shapes demonstrate that COSMOS achieves competitive performance compared to state-of-the-art SNNs in complex and noisy environments.
Authors: Zhiyi Huang, Xiaohan Shan, Jianmin Li
Abstract: Lifelong Reinforcement Learning (LRL) holds significant potential for addressing sequential tasks, but it still faces considerable challenges. A key difficulty lies in effectively preventing catastrophic forgetting and facilitating knowledge transfer while maintaining reliable decision-making performance across subsequent tasks in dynamic environments. To tackle this, we propose a novel framework, SDW (Similarity-Driven Weighting Framework), which leverages large-language-model-generated dynamic functions to precisely control the training process. The core of SDW lies in two functions pre-generated by large models: the task similarity function and the weight computation function. The task similarity function extracts multidimensional features from task descriptions to quantify the similarities and differences between tasks in terms of states, actions, and rewards. The weight computation function dynamically generates critical training parameters based on the similarity information, including the proportion of old task data stored in the Replay Buffer and the strategy consistency weight in the loss function, enabling an adaptive balance between learning new tasks and transferring knowledge from previous tasks. By generating function code offline prior to training, rather than relying on large-model inference during the training process, the SDW framework reduces computational overhead while maintaining efficiency in sequential task scenarios. Experimental results on Atari and MiniHack sequential tasks demonstrate that SDW significantly outperforms existing lifelong reinforcement learning methods.
Authors: Anthony Frion (Lab-STICC_OSE, IMT Atlantique - MEE), Lucas Drumetz (IMT Atlantique - MEE, Lab-STICC_OSE), Mauro Dalla Mura (GIPSA-SIGMAPHY), Guillaume Tochon (GIPSA-SIGMAPHY), Abdeldjalil A\"issa-El-Bey (IMT Atlantique - MEE, Lab-STICC_COSYDE)
Abstract: Following the introduction of Dynamic Mode Decomposition and its numerous extensions, many neural autoencoder-based implementations of the Koopman operator have recently been proposed. This class of methods appears to be of interest for modeling dynamical systems, either through direct long-term prediction of the evolution of the state or as a powerful embedding for downstream methods. In particular, a recent line of work has developed invertible Koopman autoencoders (IKAEs), which provide an exact reconstruction of the input state thanks to their analytically invertible encoder, based on coupling layer normalizing flow models. We identify that the conservation of the dimension imposed by the normalizing flows is a limitation for the IKAE models, and thus we propose to augment the latent state with a second, non-invertible encoder network. This results in our new model: the Augmented Invertible Koopman AutoEncoder (AIKAE). We demonstrate the relevance of the AIKAE through a series of long-term time series forecasting experiments, on satellite image time series as well as on a benchmark involving predictions based on a large lookback window of observations.
Authors: Wei Hung, Shao-Hua Sun, Ping-Chun Hsieh
Abstract: Action-constrained reinforcement learning (ACRL) is a generic framework for learning control policies with zero action constraint violation, which is required by various safety-critical and resource-constrained applications. The existing ACRL methods can typically achieve favorable constraint satisfaction but at the cost of either high computational burden incurred by the quadratic programs (QP) or increased architectural complexity due to the use of sophisticated generative models. In this paper, we propose a generic and computationally efficient framework that can adapt a standard unconstrained RL method to ACRL through two modifications: (i) To enforce the action constraints, we leverage the classic acceptance-rejection method, where we treat the unconstrained policy as the proposal distribution and derive a modified policy with feasible actions. (ii) To improve the acceptance rate of the proposal distribution, we construct an augmented two-objective Markov decision process (MDP), which include additional self-loop state transitions and a penalty signal for the rejected actions. This augmented MDP incentives the learned policy to stay close to the feasible action sets. Through extensive experiments in both robot control and resource allocation domains, we demonstrate that the proposed framework enjoys faster training progress, better constraint satisfaction, and a lower action inference time simultaneously than the state-of-the-art ACRL methods. We have made the source code publicly available to encourage further research in this direction.
Authors: Eliot Beyler (SIERRA), Francis Bach (SIERRA)
Abstract: Score-based generative models achieve state-of-the-art sampling performance by denoising a distribution perturbed by Gaussian noise. In this paper, we focus on a single deterministic denoising step, and compare the optimal denoiser for the quadratic loss, we name ''full-denoising'', to the alternative ''half-denoising'' introduced by Hyv{\"a}rinen (2024). We show that looking at the performances in term of distance between distribution tells a more nuanced story, with different assumptions on the data leading to very different conclusions.We prove that half-denoising is better than full-denoising for regular enough densities, while full-denoising is better for singular densities such as mixtures of Dirac measures or densities supported on a low-dimensional subspace. In the latter case, we prove that full-denoising can alleviate the curse of dimensionality under a linear manifold hypothesis.
Authors: Yang Ji, Ying Sun, Hengshu Zhu
Abstract: In the era of the knowledge economy, understanding how job skills influence salary is crucial for promoting recruitment with competitive salary systems and aligned salary expectations. Despite efforts on salary prediction based on job positions and talent demographics, there still lacks methods to effectively discern the set-structured skills' intricate composition effect on job salary. While recent advances in neural networks have significantly improved accurate set-based quantitative modeling, their lack of explainability hinders obtaining insights into the skills' composition effects. Indeed, model explanation for set data is challenging due to the combinatorial nature, rich semantics, and unique format. To this end, in this paper, we propose a novel intrinsically explainable set-based neural prototyping approach, namely \textbf{LGDESetNet}, for explainable salary prediction that can reveal disentangled skill sets that impact salary from both local and global perspectives. Specifically, we propose a skill graph-enhanced disentangled discrete subset selection layer to identify multi-faceted influential input subsets with varied semantics. Furthermore, we propose a set-oriented prototype learning method to extract globally influential prototypical sets. The resulting output is transparently derived from the semantic interplay between these input subsets and global prototypes. Extensive experiments on four real-world datasets demonstrate that our method achieves superior performance than state-of-the-art baselines in salary prediction while providing explainable insights into salary-influencing patterns.
Authors: Leo Zanotti
Abstract: It is shown that any continuous piecewise affine (CPA) function $\mathbb{R}^2\to\mathbb{R}$ with $p$ pieces can be represented by a ReLU neural network with two hidden layers and $O(p)$ neurons. Unlike prior work, which focused on convex pieces, this analysis considers CPA functions with connected but potentially non-convex pieces.
Authors: David E. Hernandez, Jose Ramon Chang, Torbj\"orn E. M. Nordling
Abstract: Efficient deployment of deep neural networks on resource-constrained devices demands advanced compression techniques that preserve accuracy and interoperability. This paper proposes a machine learning framework that augments Knowledge Distillation (KD) with Integrated Gradients (IG), an attribution method, to optimise the compression of convolutional neural networks. We introduce a novel data augmentation strategy where IG maps, precomputed from a teacher model, are overlaid onto training images to guide a compact student model toward critical feature representations. This approach leverages the teacher's decision-making insights, enhancing the student's ability to replicate complex patterns with reduced parameters. Experiments on CIFAR-10 demonstrate the efficacy of our method: a student model, compressed 4.1-fold from the MobileNet-V2 teacher, achieves 92.5% classification accuracy, surpassing the baseline student's 91.4% and traditional KD approaches, while reducing inference latency from 140 ms to 13 ms--a tenfold speedup. We perform hyperparameter optimisation for efficient learning. Comprehensive ablation studies dissect the contributions of KD and IG, revealing synergistic effects that boost both performance and model explainability. Our method's emphasis on feature-level guidance via IG distinguishes it from conventional KD, offering a data-driven solution for mining transferable knowledge in neural architectures. This work contributes to machine learning by providing a scalable, interpretable compression technique, ideal for edge computing applications where efficiency and transparency are paramount.
Authors: Kai Uwe Barthel, Florian Barthel, Peter Eisert
Abstract: Sorting and permutation learning are key concepts in optimization and machine learning, especially when organizing high-dimensional data into meaningful spatial layouts. The Gumbel-Sinkhorn method, while effective, requires N*N parameters to determine a full permutation matrix, making it computationally expensive for large datasets. Low-rank matrix factorization approximations reduce memory requirements to 2MN (with M << N), but they still struggle with very large problems. SoftSort, by providing a continuous relaxation of the argsort operator, allows differentiable 1D sorting, but it faces challenges with multidimensional data and complex permutations. In this paper, we present a novel method for learning permutations using only N parameters, which dramatically reduces storage costs. Our approach builds on SoftSort, but extends it by iteratively shuffling the N indices of the elements to be sorted through a separable learning process. This modification significantly improves sorting quality, especially for multidimensional data and complex optimization criteria, and outperforms pure SoftSort. Our method offers improved memory efficiency and scalability compared to existing approaches, while maintaining high-quality permutation learning. Its dramatically reduced memory requirements make it particularly well-suited for large-scale optimization tasks, such as "Self-Organizing Gaussians", where efficient and scalable permutation learning is critical.
Authors: Robin Zbinden, Nina van Tiel, Gencer Sumbul, Chiara Vanalli, Benjamin Kellenberger, Devis Tuia
Abstract: Species Distribution Models (SDMs) play a vital role in biodiversity research, conservation planning, and ecological niche modeling by predicting species distributions based on environmental conditions. The selection of predictors is crucial, strongly impacting both model accuracy and how well the predictions reflect ecological patterns. To ensure meaningful insights, input variables must be carefully chosen to match the study objectives and the ecological requirements of the target species. However, existing SDMs, including both traditional and deep learning-based approaches, often lack key capabilities for variable selection: (i) flexibility to choose relevant predictors at inference without retraining; (ii) robustness to handle missing predictor values without compromising accuracy; and (iii) explainability to interpret and accurately quantify each predictor's contribution. To overcome these limitations, we introduce MaskSDM, a novel deep learning-based SDM that enables flexible predictor selection by employing a masked training strategy. This approach allows the model to make predictions with arbitrary subsets of input variables while remaining robust to missing data. It also provides a clearer understanding of how adding or removing a given predictor affects model performance and predictions. Additionally, MaskSDM leverages Shapley values for precise predictor contribution assessments, improving upon traditional approximations. We evaluate MaskSDM on the global sPlotOpen dataset, modeling the distributions of 12,738 plant species. Our results show that MaskSDM outperforms imputation-based methods and approximates models trained on specific subsets of variables. These findings underscore MaskSDM's potential to increase the applicability and adoption of SDMs, laying the groundwork for developing foundation models in SDMs that can be readily applied to diverse ecological applications.
Authors: Amir Baghi, Jens Sj\"olund, Joakim Bergdahl, Linus Gissl\'en, Alessandro Sestini
Abstract: Multi-agent reinforcement learning has shown promise in learning cooperative behaviors in team-based environments. However, such methods often demand extensive training time. For instance, the state-of-the-art method TiZero takes 40 days to train high-quality policies for a football environment. In this paper, we hypothesize that better exploration mechanisms can improve the sample efficiency of multi-agent methods. We propose two different approaches for better exploration in TiZero: a self-supervised intrinsic reward and a random network distillation bonus. Additionally, we introduce architectural modifications to the original algorithm to enhance TiZero's computational efficiency. We evaluate the sample efficiency of these approaches through extensive experiments. Our results show that random network distillation improves training sample efficiency by 18.8% compared to the original TiZero. Furthermore, we evaluate the qualitative behavior of the models produced by both variants against a heuristic AI, with the self-supervised reward encouraging possession and random network distillation leading to a more offensive performance. Our results highlights the applicability of our random network distillation variant in practical settings. Lastly, due to the nature of the proposed method, we acknowledge its use beyond football simulation, especially in environments with strong multi-agent and strategic aspects.
Authors: Gabriele Sanguin, Arjun Pakrashi, Marco Viola, Francesco Rinaldi
Abstract: Handling uncertainty is critical for ensuring reliable decision-making in intelligent systems. Modern neural networks are known to be poorly calibrated, resulting in predicted confidence scores that are difficult to use. This article explores improving confidence estimation and calibration through the application of bilevel optimization, a framework designed to solve hierarchical problems with interdependent optimization levels. A self-calibrating bilevel neural-network training approach is introduced to improve a model's predicted confidence scores. The effectiveness of the proposed framework is analyzed using toy datasets, such as Blobs and Spirals, as well as more practical simulated datasets, such as Blood Alcohol Concentration (BAC). It is compared with a well-known and widely used calibration strategy, isotonic regression. The reported experimental results reveal that the proposed bilevel optimization approach reduces the calibration error while preserving accuracy.
Authors: Chandan Tankala, Dheeraj M. Nagaraj, Anant Raj
Abstract: Gradient flow in the 2-Wasserstein space is widely used to optimize functionals over probability distributions and is typically implemented using an interacting particle system with $n$ particles. Analyzing these algorithms requires showing (a) that the finite-particle system converges and/or (b) that the resultant empirical distribution of the particles closely approximates the optimal distribution (i.e., propagation of chaos). However, establishing efficient sufficient conditions can be challenging, as the finite particle system may produce heavily dependent random variables. In this work, we study the virtual particle stochastic approximation, originally introduced for Stein Variational Gradient Descent. This method can be viewed as a form of stochastic gradient descent in the Wasserstein space and can be implemented efficiently. In popular settings, we demonstrate that our algorithm's output converges to the optimal distribution under conditions similar to those for the infinite particle limit, and it produces i.i.d. samples without the need to explicitly establish propagation of chaos bounds.
Authors: Entao Yang, Xiaotian Zhang, Yue Shang, Ge Zhang
Abstract: While the 2024 Nobel Prize in Physics ignites a worldwide discussion on the origins of neural networks and their foundational links to physics, modern machine learning research predominantly focuses on computational and algorithmic advancements, overlooking a picture of physics. Here we introduce the concept of entropy into neural networks by reconceptualizing them as hypothetical physical systems where each parameter is a non-interacting 'particle' within a one-dimensional space. By employing a Wang-Landau algorithms, we construct the neural networks' (with up to 1 million parameters) entropy landscapes as functions of training loss and test accuracy (or loss) across four distinct machine learning tasks, including arithmetic question, real-world tabular data, image recognition, and language modeling. Our results reveal the existence of \textit{entropy advantage}, where the high-entropy states generally outperform the states reached via classical training optimizer like stochastic gradient descent. We also find this advantage is more pronounced in narrower networks, indicating a need of different training optimizers tailored to different sizes of neural networks.
Authors: Bernd Zimmering, Cec\'ilia Coelho, Vaibhav Gupta, Maria Maleshkova, Oliver Niggemann
Abstract: Modelling forced dynamical systems - where an external input drives the system state - is critical across diverse domains such as engineering, finance, and the natural sciences. In this work, we propose Laplace-Net, a decoupled, solver-free neural framework for learning forced and delay-aware systems. It leverages a Laplace transform-based approach to decompose internal dynamics, external inputs, and initial values into established theoretical concepts, enhancing interpretability. Laplace-Net promotes transferability since the system can be rapidly re-trained or fine-tuned for new forcing signals, providing flexibility in applications ranging from controller adaptation to long-horizon forecasting. Experimental results on eight benchmark datasets - including linear, non-linear, and delayed systems - demonstrate the method's improved accuracy and robustness compared to state-of-the-art approaches, particularly in handling complex and previously unseen inputs.
Authors: Nicolas Espinosa-Dice, Sanjiban Choudhury, Wen Sun, Gokul Swamy
Abstract: Interactive imitation learning (IL) is a powerful paradigm for learning to make sequences of decisions from an expert demonstrating how to perform a task. Prior work in efficient imitation learning has focused on the realizable setting, where the expert's policy lies within the learner's policy class (i.e. the learner can perfectly imitate the expert in all states). However, in practice, perfect imitation of the expert is often impossible due to differences in state information and action space expressiveness (e.g. morphological differences between robots and humans.) In this paper, we consider the more general misspecified setting, where no assumptions are made about the expert policy's realizability. We introduce a novel structural condition, reward-agnostic policy completeness, and prove that it is sufficient for interactive IL algorithms to efficiently avoid the quadratically compounding errors that stymie offline approaches like behavioral cloning. We address an additional practical constraint-the case of limited expert data-and propose a principled method for using additional offline data to further improve the sample-efficiency of interactive IL algorithms. Finally, we empirically investigate the optimal reset distribution in efficient IL under misspecification with a suite of continuous control tasks.
Authors: Ori Peleg, Natalie Lang, Stefano Rini, Nir Shlezinger, Kobi Cohen
Abstract: Federated learning (FL) enables multiple edge devices to collaboratively train a machine learning model without the need to share potentially private data. Federated learning proceeds through iterative exchanges of model updates, which pose two key challenges: First, the accumulation of privacy leakage over time, and second, communication latency. These two limitations are typically addressed separately: The former via perturbed updates to enhance privacy and the latter using user selection to mitigate latency - both at the expense of accuracy. In this work, we propose a method that jointly addresses the accumulation of privacy leakage and communication latency via active user selection, aiming to improve the trade-off among privacy, latency, and model performance. To achieve this, we construct a reward function that accounts for these three objectives. Building on this reward, we propose a multi-armed bandit (MAB)-based algorithm, termed Privacy-aware Active User SElection (PAUSE) which dynamically selects a subset of users each round while ensuring bounded overall privacy leakage. We establish a theoretical analysis, systematically showing that the reward growth rate of PAUSE follows that of the best-known rate in MAB literature. To address the complexity overhead of active user selection, we propose a simulated annealing-based relaxation of PAUSE and analyze its ability to approximate the reward-maximizing policy under reduced complexity. We numerically validate the privacy leakage, associated improved latency, and accuracy gains of our methods for the federated training in various scenarios.
Authors: Jungwon Seo, Ferhat Ozgur Catak, Chunming Rong, Kibeom Hong, Minhoe Kim
Abstract: Multi-source information fusion (MSIF) leverages diverse data streams to enhance decision-making, situational awareness, and system resilience. Federated Learning (FL) enables MSIF while preserving privacy but suffers from client drift under high data heterogeneity, leading to performance degradation. Traditional mitigation strategies rely on reference-based gradient adjustments, which can be unstable in partial participation settings. To address this, we propose Gradient Centralized Federated Learning (GC-Fed), a reference-free gradient correction method inspired by Gradient Centralization (GC). We introduce Local GC and Global GC, applying GC during local training and global aggregation, respectively. Our hybrid GC-Fed approach selectively applies GC at the feature extraction layer locally and at the classifier layer globally, improving training stability and model performance. Theoretical analysis and empirical results demonstrate that GC-Fed mitigates client drift and achieves state-of-the-art accuracy gains of up to 20% in heterogeneous settings.
Authors: Haoqi Huang, Ping Wang, Jianhua Pei, Jiacheng Wang, Shahen Alexanian, Dusit Niyato
Abstract: The rapid expansion of data from diverse sources has made anomaly detection (AD) increasingly essential for identifying unexpected observations that may signal system failures, security breaches, or fraud. As datasets become more complex and high-dimensional, traditional detection methods struggle to effectively capture intricate patterns. Advances in deep learning have made AD methods more powerful and adaptable, improving their ability to handle high-dimensional and unstructured data. This survey provides a comprehensive review of over 180 recent studies, focusing on deep learning-based AD techniques. We categorize and analyze these methods into reconstruction-based and prediction-based approaches, highlighting their effectiveness in modeling complex data distributions. Additionally, we explore the integration of traditional and deep learning methods, highlighting how hybrid approaches combine the interpretability of traditional techniques with the flexibility of deep learning to enhance detection accuracy and model transparency. Finally, we identify open issues and propose future research directions to advance the field of AD. This review bridges gaps in existing literature and serves as a valuable resource for researchers and practitioners seeking to enhance AD techniques using deep learning.
Authors: Yiman Bao, Jie Gao, Jinke He, Frans A. Oliehoek, Oded Cats
Abstract: Efficient timing in ride-matching is crucial for improving the performance of ride-hailing and ride-pooling services, as it determines the number of drivers and passengers considered in each matching process. Traditional batched matching methods often use fixed time intervals to accumulate ride requests before assigning matches. While this approach increases the number of available drivers and passengers for matching, it fails to adapt to real-time supply-demand fluctuations, often leading to longer passenger wait times and driver idle periods. To address this limitation, we propose an adaptive ride-matching strategy using deep reinforcement learning (RL) to dynamically determine when to perform matches based on real-time system conditions. Unlike fixed-interval approaches, our method continuously evaluates system states and executes matching at moments that minimize total passenger wait time. Additionally, we incorporate a potential-based reward shaping (PBRS) mechanism to mitigate sparse rewards, accelerating RL training and improving decision quality. Extensive empirical evaluations using a realistic simulator trained on real-world data demonstrate that our approach outperforms fixed-interval matching strategies, significantly reducing passenger waiting times and detour delays, thereby enhancing the overall efficiency of ride-hailing and ride-pooling systems.
Authors: Mina Kamao, Hayato Ono, Ayumu Yamashita, Kaoru Amano, Masataka Sawayama
Abstract: Alignment between human brain networks and artificial models is actively studied in machine learning and neuroscience. A widely adopted approach to explore their functional alignment is to identify metamers for both humans and models. Metamers refer to input stimuli that are physically different but equivalent within a given system. If a model's metameric space completely matched the human metameric space, the model would achieve functional alignment with humans. However, conventional methods lack direct ways to search for human metamers. Instead, researchers first develop biologically inspired models and then infer about human metamers indirectly by testing whether model metamers also appear as metamers to humans. Here, we propose the Multidimensional Adaptive Metamer Exploration (MAME) framework, enabling direct high-dimensional exploration of human metameric space. MAME leverages online image generation guided by human perceptual feedback. Specifically, it modulates reference images across multiple dimensions by leveraging hierarchical responses from convolutional neural networks (CNNs). Generated images are presented to participants whose perceptual discriminability is assessed in a behavioral task. Based on participants' responses, subsequent image generation parameters are adaptively updated online. Using our MAME framework, we successfully measured a human metameric space of over fifty dimensions within a single experiment. Experimental results showed that human discrimination sensitivity was lower for metameric images based on low-level features compared to high-level features, which image contrast metrics could not explain. The finding suggests that the model computes low-level information not essential for human perception. Our framework has the potential to contribute to developing interpretable AI and understanding of brain function in neuroscience.
Authors: Yijie Liu, Xinyi Shang, Yiqun Zhang, Yang Lu, Chen Gong, Jing-Hao Xue, Hanzi Wang
Abstract: Federated Semi-Supervised Learning (FSSL) aims to leverage unlabeled data across clients with limited labeled data to train a global model with strong generalization ability. Most FSSL methods rely on consistency regularization with pseudo-labels, converting predictions from local or global models into hard pseudo-labels as supervisory signals. However, we discover that the quality of pseudo-label is largely deteriorated by data heterogeneity, an intrinsic facet of federated learning. In this paper, we study the problem of FSSL in-depth and show that (1) heterogeneity exacerbates pseudo-label mismatches, further degrading model performance and convergence, and (2) local and global models' predictive tendencies diverge as heterogeneity increases. Motivated by these findings, we propose a simple and effective method called Semi-supervised Aggregation for Globally-Enhanced Ensemble (SAGE), that can flexibly correct pseudo-labels based on confidence discrepancies. This strategy effectively mitigates performance degradation caused by incorrect pseudo-labels and enhances consensus between local and global models. Experimental results demonstrate that SAGE outperforms existing FSSL methods in both performance and convergence. Our code is available at https://github.com/Jay-Codeman/SAGE
Authors: Ihab Asaad, Maha Shadaydeh, Joachim Denzler
Abstract: Machine learning classification models trained with empirical risk minimization (ERM) often inadvertently rely on spurious correlations. When absent in the test data, these unintended associations between non-target attributes and target labels lead to poor generalization. This paper addresses this problem from a model optimization perspective and proposes a novel method, Gradient Extrapolation for Debiased Representation Learning (GERNE), designed to learn debiased representations in both known and unknown attribute training cases. GERNE uses two distinct batches with different amounts of spurious correlations to define the target gradient as the linear extrapolation of two gradients computed from each batch's loss. It is demonstrated that the extrapolated gradient, if directed toward the gradient of the batch with fewer amount of spurious correlation, can guide the training process toward learning a debiased model. GERNE can serve as a general framework for debiasing with methods, such as ERM, reweighting, and resampling, being shown as special cases. The theoretical upper and lower bounds of the extrapolation factor are derived to ensure convergence. By adjusting this factor, GERNE can be adapted to maximize the Group-Balanced Accuracy (GBA) or the Worst-Group Accuracy. The proposed approach is validated on five vision and one NLP benchmarks, demonstrating competitive and often superior performance compared to state-of-the-art baseline methods.
Authors: Guoyou Sun, Panagiotis Karras, Qi Zhang
Abstract: Semantic communication has emerged as a promising paradigm to tackle the challenges of massive growing data traffic and sustainable data communication. It shifts the focus from data fidelity to goal-oriented or task-oriented semantic transmission. While deep learning-based methods are commonly used for semantic encoding and decoding, they struggle with the sequential nature of time series data and high computation cost, particularly in resource-constrained IoT environments. Data compression plays a crucial role in reducing transmission and storage costs, yet traditional data compression methods fall short of the demands of goal-oriented communication systems. In this paper, we propose a novel method for direct analytics on time series data compressed by the SHRINK compression algorithm. Through experimentation using outlier detection as a case study, we show that our method outperforms baselines running on uncompressed data in multiple cases, with merely 1% difference in the worst case. Additionally, it achieves four times lower runtime on average and accesses approximately 10% of the data volume, which enables edge analytics with limited storage and computation power. These results demonstrate that our approach offers reliable, high-speed outlier detection analytics for diverse IoT applications while extracting semantics from time-series data, achieving high compression, and reducing data transmission.
Authors: Chengen Wang, Murat Kantarcioglu
Abstract: In recent years, numerous graph generative models (GGMs) have been proposed. However, evaluating these models remains a considerable challenge, primarily due to the difficulty in extracting meaningful graph features that accurately represent real-world graphs. The traditional evaluation techniques, which rely on graph statistical properties like node degree distribution, clustering coefficients, or Laplacian spectrum, overlook node features and lack scalability. There are newly proposed deep learning-based methods employing graph random neural networks or contrastive learning to extract graph features, demonstrating superior performance compared to traditional statistical methods, but their experimental results also demonstrate that these methods do not always working well across different metrics. Although there are overlaps among these metrics, they are generally not interchangeable, each evaluating generative models from a different perspective. In this paper, we propose a novel method that leverages graph masked autoencoders to effectively extract graph features for GGM evaluations. We conduct extensive experiments on graphs and empirically demonstrate that our method can be more reliable and effective than previously proposed methods across a number of GGM evaluation metrics, such as "Fr\'echet Distance (FD)" and "MMD Linear". However, no single method stands out consistently across all metrics and datasets. Therefore, this study also aims to raise awareness of the significance and challenges associated with GGM evaluation techniques, especially in light of recent advances in generative models.
Authors: Fangzhi Xu, Hang Yan, Chang Ma, Haiteng Zhao, Jun Liu, Qika Lin, Zhiyong Wu
Abstract: Inference-time optimization scales computation to derive deliberate reasoning steps for effective performance. While previous search-based strategies address the short-sightedness of auto-regressive generation, the vast search space leads to excessive exploration and insufficient exploitation. To strike an efficient balance to derive the optimal step, we frame the decoding strategy as foresight sampling, leveraging simulated future steps to obtain globally optimal step estimation. Built on it, we propose a novel decoding strategy, named $\phi$-Decoding. To provide a precise and expressive estimation of step value, $\phi$-Decoding approximates two distributions via foresight and clustering. Sampling from the joint distribution, the optimal steps can be selected for exploitation. To support adaptive computation allocation, we propose in-width and in-depth pruning strategies, featuring a light-weight solution to achieve inference efficiency. Extensive experiments across seven benchmarks show $\phi$-Decoding outperforms strong baselines in both performance and efficiency. Additional analysis demonstrates its generalization across various LLMs and scalability across a wide range of computing budgets. The code will be released at https://github.com/xufangzhi/phi-Decoding, and the open-source PyPI package is coming soon.
Authors: Mikkel Jordahn, Jonas Vestergaard Jensen, Mikkel N. Schmidt, Michael Riis Andersen
Abstract: Bayesian Neural Networks (BNNs) often improve model calibration and predictive uncertainty quantification compared to point estimators such as maximum-a-posteriori (MAP). Similarly, deep ensembles (DEs) are also known to improve calibration, and therefore, it is natural to hypothesize that deep ensembles of BNNs (DE-BNNs) should provide even further improvements. In this work, we systematically investigate this across a number of datasets, neural network architectures, and BNN approximation methods and surprisingly find that when the ensembles grow large enough, DEs consistently outperform DE-BNNs on in-distribution data. To shine light on this observation, we conduct several sensitivity and ablation studies. Moreover, we show that even though DE-BNNs outperform DEs on out-of-distribution metrics, this comes at the cost of decreased in-distribution performance. As a final contribution, we open-source the large pool of trained models to facilitate further research on this topic.
Authors: Witold Wydma\'nski, Marek \'Smieja
Abstract: Feature selection in deep learning remains a critical challenge, particularly for high-dimensional tabular data where interpretability and computational efficiency are paramount. We present GFSNetwork, a novel neural architecture that performs differentiable feature selection through temperature-controlled Gumbel-Sigmoid sampling. Unlike traditional methods, where the user has to define the requested number of features, GFSNetwork selects it automatically during an end-to-end process. Moreover, GFSNetwork maintains constant computational overhead regardless of the number of input features. We evaluate GFSNetwork on a series of classification and regression benchmarks, where it consistently outperforms recent methods including DeepLasso, attention maps, as well as traditional feature selectors, while using significantly fewer features. Furthermore, we validate our approach on real-world metagenomic datasets, demonstrating its effectiveness in high-dimensional biological data. Concluding, our method provides a scalable solution that bridges the gap between neural network flexibility and traditional feature selection interpretability. We share our python implementation of GFSNetwork at https://github.com/wwydmanski/GFSNetwork, as well as a PyPi package (gfs_network).
Authors: Xin Dong, Rui Miao, Suyan Zhang, Shuaibing Jia, Leifeng Zhang, Yong Liang, Jianhua Zhang, Yi Zhun Zhu
Abstract: Repositioning drug-disease relationships has always been a hot field of research. However, actual cases of biologically validated drug relocation remain very limited, and existing models have not yet fully utilized the structural information of the drug. Furthermore, most repositioning models are only used to complete the relationship matrix, and their practicality is poor when dealing with drug cold start problems. This paper proposes a structure-enhanced multimodal relationship prediction model (SMRP). SMPR is based on the SMILE structure of the drug, using the Mol2VEC method to generate drug embedded representations, and learn disease embedded representations through heterogeneous network graph neural networks. Ultimately, a drug-disease relationship matrix is constructed. In addition, to reduce the difficulty of users' use, SMPR also provides a cold start interface based on structural similarity based on reposition results to simply and quickly predict drug-related diseases. The repositioning ability and cold start capability of the model are verified from multiple perspectives. While the AUC and ACUPR scores of repositioning reach 99% and 61% respectively, the AUC of cold start achieve 80%. In particular, the cold start Recall indicator can reach more than 70%, which means that SMPR is more sensitive to positive samples. Finally, case analysis is used to verify the practical value of the model and visual analysis directly demonstrates the improvement of the structure to the model. For quick use, we also provide local deployment of the model and package it into an executable program.
Authors: Beatriz Costa-Gomes, Joel Greer, Nikolai Juraschko, James Parkhurst, Jola Mirecka, Marjan Famili, Camila Rangel-Smith, Oliver Strickson, Alan Lowe, Mark Basham, Tom Burnley
Abstract: Ease of access to data, tools and models expedites scientific research. In structural biology there are now numerous open repositories of experimental and simulated datasets. Being able to easily access and utilise these is crucial for allowing researchers to make optimal use of their research effort. The tools presented here are useful for collating existing public cryoEM datasets and/or creating new synthetic cryoEM datasets to aid the development of novel data processing and interpretation algorithms. In recent years, structural biology has seen the development of a multitude of machine-learning based algorithms for aiding numerous steps in the processing and reconstruction of experimental datasets and the use of these approaches has become widespread. Developing such techniques in structural biology requires access to large datasets which can be cumbersome to curate and unwieldy to make use of. In this paper we present a suite of Python software packages which we collectively refer to as PERC (profet, EMPIARreader and CAKED). These are designed to reduce the burden which data curation places upon structural biology research. The protein structure fetcher (profet) package allows users to conveniently download and cleave sequences or structures from the Protein Data Bank or Alphafold databases. EMPIARreader allows lazy loading of Electron Microscopy Public Image Archive datasets in a machine-learning compatible structure. The Class Aggregator for Key Electron-microscopy Data (CAKED) package is designed to seamlessly facilitate the training of machine learning models on electron microscopy data, including electron-cryo-microscopy-specific data augmentation and labelling. These packages may be utilised independently or as building blocks in workflows. All are available in open source repositories and designed to be easily extensible to facilitate more advanced workflows if required.
Authors: Zhongwen Xu, Xianliang Wang, Siyi Li, Tao Yu, Liang Wang, Qiang Fu, Wei Yang
Abstract: We present PORTAL, a novel framework for developing artificial intelligence agents capable of playing thousands of 3D video games through language-guided policy generation. By transforming decision-making problems into language modeling tasks, our approach leverages large language models (LLMs) to generate behavior trees represented in domain-specific language (DSL). This method eliminates the computational burden associated with traditional reinforcement learning approaches while preserving strategic depth and rapid adaptability. Our framework introduces a hybrid policy structure that combines rule-based nodes with neural network components, enabling both high-level strategic reasoning and precise low-level control. A dual-feedback mechanism incorporating quantitative game metrics and vision-language model analysis facilitates iterative policy improvement at both tactical and strategic levels. The resulting policies are instantaneously deployable, human-interpretable, and capable of generalizing across diverse gaming environments. Experimental results demonstrate PORTAL's effectiveness across thousands of first-person shooter (FPS) games, showcasing significant improvements in development efficiency, policy generalization, and behavior diversity compared to traditional approaches. PORTAL represents a significant advancement in game AI development, offering a practical solution for creating sophisticated agents that can operate across thousands of commercial video games with minimal development overhead. Experiment results on the 3D video games are best viewed on https://zhongwen.one/projects/portal .
Authors: Ricardo N. Ferreira, Cl\'audia Soares
Abstract: Constrained Online Convex Optimization (COCO) can be seen as a generalization of the standard Online Convex Optimization (OCO) framework. At each round, a cost function and constraint function are revealed after a learner chooses an action. The goal is to minimize both the regret and cumulative constraint violation (CCV) against an adaptive adversary. We show for the first time that is possible to obtain the optimal $O(\sqrt{T})$ bound on both regret and CCV, improving the best known bounds of $O \left( \sqrt{T} \right)$ and $\~{O} \left( \sqrt{T} \right)$ for the regret and CCV, respectively.
Authors: Xulin Fan, Heting Gao, Ziyi Chen, Peng Chang, Mei Han, Mark Hasegawa-Johnson
Abstract: Talking head synthesis, also known as speech-to-lip synthesis, reconstructs the facial motions that align with the given audio tracks. The synthesized videos are evaluated on mainly two aspects, lip-speech synchronization and image fidelity. Recent studies demonstrate that GAN-based and diffusion-based models achieve state-of-the-art (SOTA) performance on this task, with diffusion-based models achieving superior image fidelity but experiencing lower synchronization compared to their GAN-based counterparts. To this end, we propose SyncDiff, a simple yet effective approach to improve diffusion-based models using a temporal pose frame with information bottleneck and facial-informative audio features extracted from AVHuBERT, as conditioning input into the diffusion process. We evaluate SyncDiff on two canonical talking head datasets, LRS2 and LRS3 for direct comparison with other SOTA models. Experiments on LRS2/LRS3 datasets show that SyncDiff achieves a synchronization score 27.7%/62.3% relatively higher than previous diffusion-based methods, while preserving their high-fidelity characteristics.
Authors: Cheng-Hsi Hsiao, Ellen Rathje, Krishna Kumar
Abstract: This study proposes an autoencoder approach to extract latent features from cone penetration test profiles to evaluate the potential of incorporating CPT data in an AI model. We employ autoencoders to compress 200 CPT profiles of soil behavior type index (Ic) and normalized cone resistance (qc1Ncs) into ten latent features while preserving critical information. We then utilize the extracted latent features with site parameters to train XGBoost models for predicting lateral spreading occurrences in the 2011 Christchurch earthquake. Models using the latent CPT features outperformed models with conventional CPT metrics or no CPT data, achieving over 83% accuracy. Explainable AI revealed the most crucial latent feature corresponding to soil behavior between 1-3 meter depths, highlighting this depth range's criticality for liquefaction evaluation. The autoencoder approach provides an automated technique for condensing CPT profiles into informative latent features for machine-learning liquefaction models.
Authors: Kevin Vora, Yu Zhang
Abstract: In this paper, we propose a new solution to reward adaptation (RA), the problem where the learning agent adapts to a target reward function based on one or multiple existing behaviors learned a priori under the same domain dynamics but different reward functions. Learning the target behavior from scratch is possible but often inefficient given the available source behaviors. Our work represents a new approach to RA via the manipulation of Q-functions. Assuming that the target reward function is a known function of the source reward functions, our approach to RA computes bounds of the Q function. We introduce an iterative process to tighten the bounds, similar to value iteration. This enables action pruning in the target domain before learning even starts. We refer to such a method as Q-Manipulation (Q-M). We formally prove that our pruning strategy does not affect the optimality of the returned policy while empirically show that it improves the sample complexity. Q-M is evaluated in a variety of synthetic and simulation domains to demonstrate its effectiveness, generalizability, and practicality.
Authors: Maximilian Beck, Korbinian P\"oppel, Phillip Lippe, Richard Kurle, Patrick M. Blies, G\"unter Klambauer, Sebastian B\"ock, Sepp Hochreiter
Abstract: Recent breakthroughs in solving reasoning, math and coding problems with Large Language Models (LLMs) have been enabled by investing substantial computation budgets at inference time. Therefore, inference speed is one of the most critical properties of LLM architectures, and there is a growing need for LLMs that are efficient and fast at inference. Recently, LLMs built on the xLSTM architecture have emerged as a powerful alternative to Transformers, offering linear compute scaling with sequence length and constant memory usage, both highly desirable properties for efficient inference. However, such xLSTM-based LLMs have yet to be scaled to larger models and assessed and compared with respect to inference speed and efficiency. In this work, we introduce xLSTM 7B, a 7-billion-parameter LLM that combines xLSTM's architectural benefits with targeted optimizations for fast and efficient inference. Our experiments demonstrate that xLSTM 7B achieves performance on downstream tasks comparable to other similar-sized LLMs, while providing significantly faster inference speeds and greater efficiency compared to Llama- and Mamba-based LLMs. These results establish xLSTM 7B as the fastest and most efficient 7B LLM, offering a solution for tasks that require large amounts of test-time computation. Our work highlights xLSTM's potential as a foundational architecture for methods building on heavy use of LLM inference. Our model weights, model code and training code are open-source.
Authors: Vincent Herrmann, R\'obert Csord\'as, J\"urgen Schmidhuber
Abstract: Detecting when a neural sequence model does "interesting" computation is an open problem. The next token prediction loss is a poor indicator: Low loss can stem from trivially predictable sequences that are uninteresting, while high loss may reflect unpredictable but also irrelevant information that can be ignored by the model. We propose a better metric: measuring the model's ability to predict its own future hidden states. We show empirically that this metric -- in contrast to the next token prediction loss -- correlates with the intuitive interestingness of the task. To measure predictability, we introduce the architecture-agnostic "prediction of hidden states" (PHi) layer that serves as an information bottleneck on the main pathway of the network (e.g., the residual stream in Transformers). We propose a novel learned predictive prior that enables us to measure the novel information gained in each computation step, which serves as our metric. We show empirically that our metric predicts the description length of formal languages learned in-context, the complexity of mathematical reasoning problems, and the correctness of self-generated reasoning chains.
Authors: Marta Grzeskiewicz
Abstract: Determining consumer preferences and utility is a foundational challenge in economics. They are central in determining consumer behaviour through the utility-maximising consumer decision-making process. However, preferences and utilities are not observable and may not even be known to the individual making the choice; only the outcome is observed in the form of demand. Without the ability to observe the decision-making mechanism, demand estimation becomes a challenging task and current methods fall short due to lack of scalability or ability to identify causal effects. Estimating these effects is critical when considering changes in policy, such as pricing, the impact of taxes and subsidies, and the effect of a tariff. To address the shortcomings of existing methods, we combine revealed preference theory and inverse reinforcement learning to present a novel algorithm, Preference Extraction and Reward Learning (PEARL) which, to the best of our knowledge, is the only algorithm that can uncover a representation of the utility function that best rationalises observed consumer choice data given a specified functional form. We introduce a flexible utility function, the Input-Concave Neural Network which captures complex relationships across goods, including cross-price elasticities. Results show PEARL outperforms the benchmark on both noise-free and noisy synthetic data.
Authors: Giacomo Arcieri, Konstantinos G. Papakonstantinou, Daniel Straub, Eleni Chatzi
Abstract: This work introduces a novel deep learning-based architecture, termed the Deep Belief Markov Model (DBMM), which provides efficient, model-formulation agnostic inference in Partially Observable Markov Decision Process (POMDP) problems. The POMDP framework allows for modeling and solving sequential decision-making problems under observation uncertainty. In complex, high-dimensional, partially observable environments, existing methods for inference based on exact computations (e.g., via Bayes' theorem) or sampling algorithms do not scale well. Furthermore, ground truth states may not be available for learning the exact transition dynamics. DBMMs extend deep Markov models into the partially observable decision-making framework and allow efficient belief inference entirely based on available observation data via variational inference methods. By leveraging the potency of neural networks, DBMMs can infer and simulate non-linear relationships in the system dynamics and naturally scale to problems with high dimensionality and discrete or continuous variables. In addition, neural network parameters can be dynamically updated efficiently based on data availability. DBMMs can thus be used to infer a belief variable, thus enabling the derivation of POMDP solutions over the belief space. We evaluate the efficacy of the proposed methodology by evaluating the capability of model-formulation agnostic inference of DBMMs in benchmark problems that include discrete and continuous variables.
Authors: Runzhi Wang, Prianka Sengupta, Yiran Chen, Jiang Hu
Abstract: In chip design planning, obtaining reliable performance and power forecasts for various design options is of critical importance. Traditionally, this involves using system-level models, which often lack accuracy, or trial synthesis, which is both labor-intensive and time-consuming. We introduce a new methodology, called Lorecast, which accepts English prompts as input to rapidly generate layout-aware performance and power estimates. This approach bypasses the need for HDL code development or synthesis, making it both fast and user-friendly. Experimental results demonstrate that Lorecast achieves accuracy within a few percent of error compared to post-layout analysis.
Authors: Abhishek Moitra, Arkapravo Ghosh, Shrey Agarwal, Aporva Amarnath, Karthik Swaminathan, Priyadarshini Panda
Abstract: The computational and memory challenges of large language models (LLMs) have sparked several optimization approaches towards their efficient implementation. While prior LLM-targeted quantization, and prior works on sparse acceleration have significantly mitigated the memory and computation bottleneck, they do so assuming high power platforms such as GPUs and server-class FPGAs with large off-chip memory bandwidths and employ a generalized matrix multiplication (GEMM) execution of all the layers in the decoder. In such a GEMM-based execution, data is fetched from an off-chip memory, computed and stored back. However, at reduced off-chip memory capacities, as is the case with low-power edge devices, this implementation strategy significantly increases the attention computation latency owing to the repeated storage and fetch of large intermediate tokens to and from the off-chip memory. Moreover, fetching the weight matrices from a bandwidth constrained memory further aggravates the memory bottleneck problem. To this end, we introduce MEADOW, a framework that significantly reduces the off-chip memory access for LLMs with a novel token-parallel head-sequential (TPHS) dataflow. Additionally, MEADOW applies weight packing that performs loss-less decomposition of large weight matrices to their unique elements thereby, reducing the enormous weight fetch latency. MEADOW demonstrates 1.5x and 2.5x lower decode and prefill latency, respectively, compared to a GEMM-based LLM implementation on the low power Xilinx ZCU102 FPGA platform that consumes less than 10W. Additionally, MEADOW achieves an end-to-end latency improvement of over 40%, compared to prior LLM optimization works.
Authors: Micol Spitale, Srikar Babu, Serhan Cakmak, Jiaee Cheong, Hatice Gunes
Abstract: One of the primary goals of Human-Robot Interaction (HRI) research is to develop robots that can interpret human behavior and adapt their responses accordingly. Adaptive learning models, such as continual and reinforcement learning, play a crucial role in improving robots' ability to interact effectively in real-world settings. However, these models face significant challenges due to the limited availability of real-world data, particularly in sensitive domains like healthcare and well-being. This data scarcity can hinder a robot's ability to adapt to new situations. To address these challenges, causality provides a structured framework for understanding and modeling the underlying relationships between actions, events, and outcomes. By moving beyond mere pattern recognition, causality enables robots to make more explainable and generalizable decisions. This paper presents an exploratory causality-based analysis through a case study of an adaptive robotic coach delivering positive psychology exercises over four weeks in a workplace setting. The robotic coach autonomously adapts to multimodal human behaviors, such as facial valence and speech duration. By conducting both macro- and micro-level causal analyses, this study aims to gain deeper insights into how adaptability can enhance well-being during interactions. Ultimately, this research seeks to advance our understanding of how causality can help overcome challenges in HRI, particularly in real-world applications.
Authors: Christopher Bennett, Kerstin Eder
Abstract: Microelectronic design verification remains a critical bottleneck in device development, traditionally mitigated by expanding verification teams and computational resources. Since the late 1990s, machine learning (ML) has been proposed to enhance verification efficiency, yet many techniques have not achieved mainstream adoption. This review, from the perspective of verification and ML practitioners, examines the application of ML in dynamic-based techniques for functional verification of microelectronic designs, and provides a starting point for those new to this interdisciplinary field. Historical trends, techniques, ML types, and evaluation baselines are analysed to understand why previous research has not been widely adopted in industry. The review highlights the application of ML, the techniques used and critically discusses their limitations and successes. Although there is a wealth of promising research, real-world adoption is hindered by challenges in comparing techniques, identifying suitable applications, and the expertise required for implementation. This review proposes that the field can progress through the creation and use of open datasets, common benchmarks, and verification targets. By establishing open evaluation criteria, industry can guide future research. Parallels with ML in software verification suggest potential for collaboration. Additionally, greater use of open-source designs and verification environments can allow more researchers from outside the hardware verification discipline to contribute to the challenge of verifying microelectronic designs.
Authors: V\'aclav Jirkovsk\'y, Ji\v{r}\'i Kubal\'ik, Petr Kadera, Arnd Schirrmann, Andreas Mitschke, Andreas Zindel
Abstract: This paper presents a new complex optimization problem in the field of automatic design of advanced industrial systems and proposes a hybrid optimization approach to solve the problem. The problem is multi-objective as it aims at finding solutions that minimize CO2 emissions, transportation time, and costs. The optimization approach combines an evolutionary algorithm and classical mathematical programming to design resilient and sustainable global manufacturing networks. Further, it makes use of the OWL ontology for data consistency and constraint management. The experimental validation demonstrates the effectiveness of the approach in both single and double sourcing scenarios. The proposed methodology, in general, can be applied to any industry case with complex manufacturing and supply chain challenges.
Authors: E Harshith Kumar Yadav, Rahul Narava, Anshika, Shashi Shekher Jha
Abstract: Managing equal charge levels in active cell balancing while charging a Li-ion battery is challenging. An imbalance in charge levels affects the state of health of the battery, along with the concerns of thermal runaway and fire hazards. Traditional methods focus on safety assurance as a trade-off between safety and charging time. Others deal with battery-specific conditions to ensure safety, therefore losing on the generalization of the control strategies over various configurations of batteries. In this work, we propose a method to learn safe battery charging actions by using a safety-layer as an add-on over a Deep Reinforcement Learning (RL) agent. The safety layer perturbs the agent's action to prevent the battery from encountering unsafe or dangerous states. Further, our Deep RL framework focuses on learning a generalized policy that can be effectively employed with varying configurations of batteries. Our experimental results demonstrate that the safety-layer based action perturbation incurs fewer safety violations by avoiding unsafe states along with learning a robust policy for several battery configurations.
Authors: Juan C. Perdomo
Abstract: Social predictions do not passively describe the future; they actively shape it. They inform actions and change individual expectations in ways that influence the likelihood of the predicted outcome. Given these dynamics, to what extent can social events be predicted? This question was discussed throughout the 20th century by authors like Merton, Morgenstern, Simon, and others who considered it a central issue in social science methodology. In this work, we provide a modern answer to this old problem. Using recent ideas from performative prediction and outcome indistinguishability, we establish that one can always efficiently predict social events accurately, regardless of how predictions influence data. While achievable, we also show that these predictions are often undesirable, highlighting the limitations of previous desiderata. We end with a discussion of various avenues forward.
Authors: Gabriele D'Acunto, Claudio Battiloro
Abstract: Recent advances in artificial intelligence reveal the limits of purely predictive systems and call for a shift toward causal and collaborative reasoning. Drawing inspiration from the revolution of Grothendieck in mathematics, we introduce the relativity of causal knowledge, which posits structural causal models (SCMs) are inherently imperfect, subjective representations embedded within networks of relationships. By leveraging category theory, we arrange SCMs into a functor category and show that their observational and interventional probability measures naturally form convex structures. This result allows us to encode non-intervened SCMs with convex spaces of probability measures. Next, using sheaf theory, we construct the network sheaf and cosheaf of causal knowledge. These structures enable the transfer of causal knowledge across the network while incorporating interventional consistency and the perspective of the subjects, ultimately leading to the formal, mathematical definition of relative causal knowledge.
Authors: Elena Ballante, Pietro Muliere, Silvia Figini
Abstract: This paper proposes a new class of predictive models for survival analysis called Generalized Bayesian Ensemble Survival Tree (GBEST). It is well known that survival analysis poses many different challenges, in particular when applied to small data or censorship mechanism. Our contribution is the proposal of an ensemble approach that uses Bayesian bootstrap and beta Stacy bootstrap methods to improve the outcome in survival application with a special focus on small datasets. More precisely, a novel approach to integrate Beta Stacy Bayesian bootstrap in bagging tree models for censored data is proposed in this paper. Empirical evidence achieved on simulated and real data underlines that our approach performs better in terms of predictive performances and stability of the results compared with classical survival models available in the literature. In terms of methodology our novel contribution considers the adaptation of recent Bayesian ensemble approaches to survival data, providing a new model called Generalized Bayesian Ensemble Survival Tree (GBEST). A further result in terms of computational novelty is the implementation in R of GBEST, available in a public GitHub repository.
Authors: Yuhao Wang, Enlu Zhou
Abstract: In this paper, we propose a general and novel formulation of ranking and selection with the existence of streaming input data. The collection of multiple streams of such data may consume different types of resources, and hence can be conducted simultaneously. To utilize the streaming input data, we aggregate simulation outputs generated under heterogeneous input distributions over time to form a performance estimator. By characterizing the asymptotic behavior of the performance estimators, we formulate two optimization problems to optimally allocate budgets for collecting input data and running simulations. We then develop a multi-stage simultaneous budget allocation procedure and provide its statistical guarantees such as consistency and asymptotic normality. We conduct several numerical studies to demonstrate the competitive performance of the proposed procedure.
Authors: Bangzheng Li, Fei Wang, Wenxuan Zhou, Nan Xu, Ben Zhou, Sheng Zhang, Hoifung Poon, Muhao Chen
Abstract: Vision-Language Models (VLMs) leverage aligned visual encoders to transform images into visual tokens, allowing them to be processed similarly to text by the backbone large language model (LLM). This unified input paradigm enables VLMs to excel in vision-language tasks such as visual question answering (VQA). To improve fine-grained visual reasoning, recent advancements in vision-language modeling introduce image cropping techniques that feed all encoded sub-images into the model. However, this approach significantly increases the number of visual tokens, leading to inefficiency and potential distractions for the LLM. To address the generalization challenges of image representation in VLMs, we propose a lightweight, universal framework that seamlessly integrates with existing VLMs to enhance their ability to process finegrained details. Our method leverages textual semantics to identify key visual areas, improving VQA performance without requiring any retraining of the VLM. Additionally, it incorporates textual signals into the visual encoding process, enhancing both efficiency and effectiveness. The proposed method, SEMCLIP, strengthens the visual understanding of a 7B VLM, LLaVA-1.5 by 3.3% on average across 7 benchmarks, and particularly by 5.3% on the challenging detailed understanding benchmark V*.
Authors: Sanayya A, Amoolya Shetty, Abhijeet Sharma, Venkatesh Ravichandran, Masthan Wali Gosuvarapalli, Sarthak Jain, Priyamvada Nanjundiah, Ujjal Kr Dutta, Divya Sharma
Abstract: In agricultural management, precise Ground Truth (GT) data is crucial for accurate Machine Learning (ML) based crop classification. Yet, issues like crop mislabeling and incorrect land identification are common. We propose a multi-level GT cleaning framework while utilizing multi-temporal Sentinel-2 data to address these issues. Specifically, this framework utilizes generating embeddings for farmland, clustering similar crop profiles, and identification of outliers indicating GT errors. We validated clusters with False Colour Composite (FCC) checks and used distance-based metrics to scale and automate this verification process. The importance of cleaning the GT data became apparent when the models were trained on the clean and unclean data. For instance, when we trained a Random Forest model with the clean GT data, we achieved upto 70\% absolute percentage points higher for the F1 score metric. This approach advances crop classification methodologies, with potential for applications towards improving loan underwriting and agricultural decision-making.
Authors: Yiwei Chen, Yuguang Yao, Yihua Zhang, Bingquan Shen, Gaowen Liu, Sijia Liu
Abstract: Recent vision-language models (VLMs) have made remarkable strides in generative modeling with multimodal inputs, particularly text and images. However, their susceptibility to generating harmful content when exposed to unsafe queries raises critical safety concerns. While current alignment strategies primarily rely on supervised safety fine-tuning with curated datasets, we identify a fundamental limitation we call the "safety mirage" where supervised fine-tuning inadvertently reinforces spurious correlations between superficial textual patterns and safety responses, rather than fostering deep, intrinsic mitigation of harm. We show that these spurious correlations leave fine-tuned VLMs vulnerable even to a simple one-word modification-based attack, where substituting a single word in text queries with a spurious correlation-inducing alternative can effectively bypass safeguards. Additionally, these correlations contribute to the over prudence, causing fine-tuned VLMs to refuse benign queries unnecessarily. To address this issue, we show machine unlearning (MU) as a powerful alternative to supervised safety fine-tuning as it avoids biased feature-label mappings and directly removes harmful knowledge from VLMs while preserving their general capabilities. Extensive evaluations across safety benchmarks show that under one-word attacks, MU-based alignment reduces the attack success rate by up to 60.17% and cuts unnecessary rejections by over 84.20%. Codes are available at https://github.com/OPTML-Group/VLM-Safety-MU. WARNING: There exist AI generations that may be offensive in nature.
Authors: Peiqi Yang, Conglong Xu, Hao Wu
Abstract: We prove a convergence theorem for stochastic gradient descents on manifolds with adaptive learning rate and apply it to the weighted low-rank approximation problem.
Authors: Tianwei Lan, Luca Demetrio, Farid Nait-Abdesselam, Yufei Han, Simone Aonzo
Abstract: Machine learning (ML) malware detectors rely heavily on crowd-sourced AntiVirus (AV) labels, with platforms like VirusTotal serving as a trusted source of malware annotations. But what if attackers could manipulate these labels to classify benign software as malicious? We introduce label spoofing attacks, a new threat that contaminates crowd-sourced datasets by embedding minimal and undetectable malicious patterns into benign samples. These patterns coerce AV engines into misclassifying legitimate files as harmful, enabling poisoning attacks against ML-based malware classifiers trained on those data. We demonstrate this scenario by developing AndroVenom, a methodology for polluting realistic data sources, causing consequent poisoning attacks against ML malware detectors. Experiments show that not only state-of-the-art feature extractors are unable to filter such injection, but also various ML models experience Denial of Service already with 1% poisoned samples. Additionally, attackers can flip decisions of specific unaltered benign samples by modifying only 0.015% of the training data, threatening their reputation and market share and being unable to be stopped by anomaly detectors on training data. We conclude our manuscript by raising the alarm on the trustworthiness of the training process based on AV annotations, requiring further investigation on how to produce proper labels for ML malware detectors.
Authors: Nirmalya Thakur, Niven Francis Da Guia Fernandes, Madje Tobi Marc'Avent Tchona
Abstract: Long COVID continues to challenge public health by affecting a considerable number of individuals who have recovered from acute SARS-CoV-2 infection yet endure prolonged and often debilitating symptoms. Social media has emerged as a vital resource for those seeking real-time information, peer support, and validating their health concerns related to Long COVID. This paper examines recent works focusing on mining, analyzing, and interpreting user-generated content on social media platforms to capture the broader discourse on persistent post-COVID conditions. A novel transformer-based zero-shot learning approach serves as the foundation for classifying research papers in this area into four primary categories: Clinical or Symptom Characterization, Advanced NLP or Computational Methods, Policy Advocacy or Public Health Communication, and Online Communities and Social Support. This methodology achieved an average confidence of 0.7788, with the minimum and maximum confidence being 0.1566 and 0.9928, respectively. This model showcases the ability of advanced language models to categorize research papers without any training data or predefined classification labels, thus enabling a more rapid and scalable assessment of existing literature. This paper also highlights the multifaceted nature of Long COVID research by demonstrating how advanced computational techniques applied to social media conversations can reveal deeper insights into the experiences, symptoms, and narratives of individuals affected by Long COVID.
Authors: Alexander Weers, Alexander H. Berger, Laurin Lux, Peter Sch\"uffler, Daniel Rueckert, Johannes C. Paetzold
Abstract: The histopathological classification of whole-slide images (WSIs) is a fundamental task in digital pathology; yet it requires extensive time and expertise from specialists. While deep learning methods show promising results, they typically process WSIs by dividing them into artificial patches, which inherently prevents a network from learning from the entire image context, disregards natural tissue structures and compromises interpretability. Our method overcomes this limitation through a novel graph-based framework that constructs WSI graph representations. The WSI-graph efficiently captures essential histopathological information in a compact form. We build tissue representations (nodes) that follow biological boundaries rather than arbitrary patches all while providing interpretable features for explainability. Through adaptive graph coarsening guided by learned embeddings, we progressively merge regions while maintaining discriminative local features and enabling efficient global information exchange. In our method's final step, we solve the diagnostic task through a graph attention network. We empirically demonstrate strong performance on multiple challenging tasks such as cancer stage classification and survival prediction, while also identifying predictive factors using Integrated Gradients. Our implementation is publicly available at https://github.com/HistoGraph31/pix2pathology
Authors: Vitaly Feldman, Audra McMillan, Guy N. Rothblum, Kunal Talwar
Abstract: Pan-privacy was proposed by Dwork et al. as an approach to designing a private analytics system that retains its privacy properties in the face of intrusions that expose the system's internal state. Motivated by federated telemetry applications, we study local pan-privacy, where privacy should be retained under repeated unannounced intrusions on the local state. We consider the problem of monitoring the count of an event in a federated system, where event occurrences on a local device should be hidden even from an intruder on that device. We show that under reasonable constraints, the goal of providing information-theoretic differential privacy under intrusion is incompatible with collecting telemetry information. We then show that this problem can be solved in a scalable way using standard cryptographic primitives.
Authors: Jutika Borah, Hidam Kumarjit Singh
Abstract: Accurate and reliable image classification is crucial in radiology, where diagnostic decisions significantly impact patient outcomes. Conventional deep learning models tend to produce overconfident predictions despite underlying uncertainties, potentially leading to misdiagnoses. Attention mechanisms have emerged as powerful tools in deep learning, enabling models to focus on relevant parts of the input data. Combined with feature fusion, they can be effective in addressing uncertainty challenges. Cross-attention has become increasingly important in medical image analysis for capturing dependencies across features and modalities. This paper proposes a novel dual cross-attention fusion model for medical image analysis by addressing key challenges in feature integration and interpretability. Our approach introduces a bidirectional cross-attention mechanism with refined channel and spatial attention that dynamically fuses feature maps from EfficientNetB4 and ResNet34 leveraging multi-network contextual dependencies. The refined features through channel and spatial attention highlights discriminative patterns crucial for accurate classification. The proposed model achieved AUC of 99.75%, 100%, 99.93% and 98.69% and AUPR of 99.81%, 100%, 99.97%, and 96.36% on Covid-19, Tuberculosis, Pneumonia Chest X-ray images and Retinal OCT images respectively. The entropy values and several high uncertain samples give an interpretable visualization from the model enhancing transparency. By combining multi-scale feature extraction, bidirectional attention and uncertainty estimation, our proposed model strongly impacts medical image analysis.
Authors: Yue Ju, Bo Wahlberg, H\r{a}kan Hjalmarsson
Abstract: Regularized system identification has become a significant complement to more classical system identification. It has been numerically shown that kernel-based regularized estimators often perform better than the maximum likelihood estimator in terms of minimizing mean squared error (MSE). However, regularized estimators often require hyper-parameter estimation. This paper focuses on ridge regression and the regularized estimator by employing the empirical Bayes hyper-parameter estimator. We utilize the excess MSE to quantify the MSE difference between the empirical-Bayes-based regularized estimator and the maximum likelihood estimator for large sample sizes. We then exploit the excess MSE expressions to develop both a family of generalized Bayes estimators and a family of closed-form biased estimators. They have the same excess MSE as the empirical-Bayes-based regularized estimator but eliminate the need for hyper-parameter estimation. Moreover, we conduct numerical simulations to show that the performance of these new estimators is comparable to the empirical-Bayes-based regularized estimator, while computationally, they are more efficient.
Authors: Arvind Raghavan, Elias Bareinboim
Abstract: It is commonly believed that, in a real-world environment, samples can only be drawn from observational and interventional distributions, corresponding to Layers 1 and 2 of the Pearl Causal Hierarchy. Layer 3, representing counterfactual distributions, is believed to be inaccessible by definition. However, Bareinboim, Forney, and Pearl (2015) introduced a procedure that allows an agent to sample directly from a counterfactual distribution, leaving open the question of what other counterfactual quantities can be estimated directly via physical experimentation. We resolve this by introducing a formal definition of realizability, the ability to draw samples from a distribution, and then developing a complete algorithm to determine whether an arbitrary counterfactual distribution is realizable given fundamental physical constraints, such as the inability to go back in time and subject the same unit to a different experimental condition. We illustrate the implications of this new framework for counterfactual data collection using motivating examples from causal fairness and causal reinforcement learning. While the baseline approach in these motivating settings typically follows an interventional or observational strategy, we show that a counterfactual strategy provably dominates both.
Authors: Bhiman Kumar Baghel, Scott M. Jordan, Zheyuan Ryan Shi, Xiang Lorraine Li
Abstract: Large Language Models (LLMs) are used in various downstream language tasks, making it crucial to keep their knowledge up-to-date, but both retraining and fine-tuning the model can be costly. Model editing offers an efficient and effective alternative by a single update to only a key subset of model parameters. While being efficient, these methods are not perfect. Sometimes knowledge edits are unsuccessful, i.e., UnderEdit, or the edit contaminated neighboring knowledge that should remain unchanged, i.e., OverEdit. To address these limitations, we propose iterative model editing, based on our hypothesis that a single parameter update is often insufficient, to mitigate UnderEdit, and neighbor-assisted model editing, which incorporates neighboring knowledge during editing to minimize OverEdit. Extensive experiments demonstrate that our methods effectively reduce UnderEdit up to 38 percentage points and OverEdit up to 6 percentage points across multiple model editing algorithms, LLMs, and benchmark datasets.
Authors: Hilal Asi, Vitaly Feldman, Hannah Keller, Guy N. Rothblum, Kunal Talwar
Abstract: We revisit the problem of secure aggregation of high-dimensional vectors in a two-server system such as Prio. These systems are typically used to aggregate vectors such as gradients in private federated learning, where the aggregate itself is protected via noise addition to ensure differential privacy. Existing approaches require communication scaling with the dimensionality, and thus limit the dimensionality of vectors one can efficiently process in this setup. We propose PREAMBLE: Private Efficient Aggregation Mechanism for BLock-sparse Euclidean Vectors. PREAMBLE is a novel extension of distributed point functions that enables communication- and computation-efficient aggregation of block-sparse vectors, which are sparse vectors where the non-zero entries occur in a small number of clusters of consecutive coordinates. We then show that PREAMBLE can be combined with random sampling and privacy amplification by sampling results, to allow asymptotically optimal privacy-utility trade-offs for vector aggregation, at a fraction of the communication cost. When coupled with recent advances in numerical privacy accounting, our approach incurs a negligible overhead in noise variance, compared to the Gaussian mechanism used with Prio.
Authors: Kun Su, Krishna Sayana, Hubert Pham, James Pine, Yuri Vasilevski, Raghavendra Vasudeva, Marialena Kyriakidi, Liam Hebert, Ambarish Jash, Anushya Subbiah, Sukhdeep Sodhi
Abstract: This paper introduces a novel dataset REGEN (Reviews Enhanced with GEnerative Narratives), designed to benchmark the conversational capabilities of recommender Large Language Models (LLMs), addressing the limitations of existing datasets that primarily focus on sequential item prediction. REGEN extends the Amazon Product Reviews dataset by inpainting two key natural language features: (1) user critiques, representing user "steering" queries that lead to the selection of a subsequent item, and (2) narratives, rich textual outputs associated with each recommended item taking into account prior context. The narratives include product endorsements, purchase explanations, and summaries of user preferences. Further, we establish an end-to-end modeling benchmark for the task of conversational recommendation, where models are trained to generate both recommendations and corresponding narratives conditioned on user history (items and critiques). For this joint task, we introduce a modeling framework LUMEN (LLM-based Unified Multi-task Model with Critiques, Recommendations, and Narratives) which uses an LLM as a backbone for critiquing, retrieval and generation. We also evaluate the dataset's quality using standard auto-rating techniques and benchmark it by training both traditional and LLM-based recommender models. Our results demonstrate that incorporating critiques enhances recommendation quality by enabling the recommender to learn language understanding and integrate it with recommendation signals. Furthermore, LLMs trained on our dataset effectively generate both recommendations and contextual narratives, achieving performance comparable to state-of-the-art recommenders and language models.
Authors: Jingxuan Zhang, Robert J. Hart, Ziqian Bi, Shiaofen Fang, Susan Walsh
Abstract: The use of the iris as a biometric identifier has increased dramatically over the last 30 years, prompting privacy and security concerns about the use of iris images in research. It can be difficult to acquire iris image databases due to ethical concerns, and this can be a barrier for those performing biometrics research. In this paper, we describe and show how to create a database of realistic, biometrically unidentifiable colored iris images by training a diffusion model within an open-source diffusion framework. Not only were we able to verify that our model is capable of creating iris textures that are biometrically unique from the training data, but we were also able to verify that our model output creates a full distribution of realistic iris pigmentations. We highlight the fact that the utility of diffusion networks to achieve these criteria with relative ease, warrants additional research in its use within the context of iris database generation and presentation attack security.
Authors: Naresh Kumar Devulapally, Mingzhen Huang, Vishal Asnani, Shruti Agarwal, Siwei Lyu, Vishnu Suresh Lokhande
Abstract: Invisible watermarking of AI-generated images can help with copyright protection, enabling detection and identification of AI-generated media. In this work, we present a novel approach to watermark images of T2I Latent Diffusion Models (LDMs). By only fine-tuning text token embeddings $W_*$, we enable watermarking in selected objects or parts of the image, offering greater flexibility compared to traditional full-image watermarking. Our method leverages the text encoder's compatibility across various LDMs, allowing plug-and-play integration for different LDMs. Moreover, introducing the watermark early in the encoding stage improves robustness to adversarial perturbations in later stages of the pipeline. Our approach achieves $99\%$ bit accuracy ($48$ bits) with a $10^5 \times$ reduction in model parameters, enabling efficient watermarking.
Authors: Austin Shouli, Ankur Barthwal, Molly Campbell, Ajay Kumar Shrestha
Abstract: The rapid expansion of Artificial Intelligence (AI) in digital platforms used by youth has created significant challenges related to privacy, autonomy, and data protection. While AI-driven personalization offers enhanced user experiences, it often operates without clear ethical boundaries, leaving young users vulnerable to data exploitation and algorithmic biases. This paper presents a call to action for ethical AI governance, advocating for a structured framework that ensures youth-centred privacy protections, transparent data practices, and regulatory oversight. We outline key areas requiring urgent intervention, including algorithmic transparency, privacy education, parental data-sharing ethics, and accountability measures. Through this approach, we seek to empower youth with greater control over their digital identities and propose actionable strategies for policymakers, AI developers, and educators to build a fairer and more accountable AI ecosystem.
Authors: Ahcen Aliouat, Elsa Dupraz
Abstract: In the emerging field of goal-oriented communications, the focus has shifted from reconstructing data to directly performing specific learning tasks, such as classification, segmentation, or pattern recognition, on the received coded data. In the commonly studied scenario of classification from compressed images, a key objective is to enable learning directly on entropy-coded data, thereby bypassing the computationally intensive step of data reconstruction. Conventional entropy-coding methods, such as Huffman and Arithmetic coding, are effective for compression but disrupt the data structure, making them less suitable for direct learning without decoding. This paper investigates the use of low-density parity-check (LDPC) codes -- originally designed for channel coding -- as an alternative entropy-coding approach. It is hypothesized that the structured nature of LDPC codes can be leveraged more effectively by deep learning models for tasks like classification. At the receiver side, gated recurrent unit (GRU) models are trained to perform image classification directly on LDPC-coded data. Experiments on datasets like MNIST, Fashion-MNIST, and CIFAR show that LDPC codes outperform Huffman and Arithmetic coding in classification tasks, while requiring significantly smaller learning models. Furthermore, the paper analyzes why LDPC codes preserve data structure more effectively than traditional entropy-coding techniques and explores the impact of key code parameters on classification performance. These results suggest that LDPC-based entropy coding offers an optimal balance between learning efficiency and model complexity, eliminating the need for prior decoding.
Authors: Chong Su, Yingbin Fu, Zheyuan Hu, Jing Yang, Param Hanji, Shaojun Wang, Xuan Zhao, Cengiz \"Oztireli, Fangcheng Zhong
Abstract: We introduce CHOrD, a novel framework for scalable synthesis of 3D indoor scenes, designed to create house-scale, collision-free, and hierarchically structured indoor digital twins. In contrast to existing methods that directly synthesize the scene layout as a scene graph or object list, CHOrD incorporates a 2D image-based intermediate layout representation, enabling effective prevention of collision artifacts by successfully capturing them as out-of-distribution (OOD) scenarios during generation. Furthermore, unlike existing methods, CHOrD is capable of generating scene layouts that adhere to complex floor plans with multi-modal controls, enabling the creation of coherent, house-wide layouts robust to both geometric and semantic variations in room structures. Additionally, we propose a novel dataset with expanded coverage of household items and room configurations, as well as significantly improved data quality. CHOrD demonstrates state-of-the-art performance on both the 3D-FRONT and our proposed datasets, delivering photorealistic, spatially coherent indoor scene synthesis adaptable to arbitrary floor plan variations.
Authors: Jianqi Gao, Xizheng Pang, Qi Liu, Yanjie Li
Abstract: Reinforcement learning-based mapless navigation holds significant potential. However, it faces challenges in indoor environments with local minima area. This paper introduces a safe mapless navigation framework utilizing hierarchical reinforcement learning (HRL) to enhance navigation through such areas. The high-level policy creates a sub-goal to direct the navigation process. Notably, we have developed a sub-goal update mechanism that considers environment congestion, efficiently avoiding the entrapment of the robot in local minimum areas. The low-level motion planning policy, trained through safe reinforcement learning, outputs real-time control instructions based on acquired sub-goal. Specifically, to enhance the robot's environmental perception, we introduce a new obstacle encoding method that evaluates the impact of obstacles on the robot's motion planning. To validate the performance of our HRL-based navigation framework, we conduct simulations in office, home, and restaurant environments. The findings demonstrate that our HRL-based navigation framework excels in both static and dynamic scenarios. Finally, we implement the HRL-based navigation framework on a TurtleBot3 robot for physical validation experiments, which exhibits its strong generalization capabilities.
Authors: Kaining Shi, Cong Ma
Abstract: This paper introduces a novel theoretical framework for auditing differential privacy (DP) in a black-box setting. Leveraging the concept of $f$-differential privacy, we explicitly define type I and type II errors and propose an auditing mechanism based on conformal inference. Our approach robustly controls the type I error rate under minimal assumptions. Furthermore, we establish a fundamental impossibility result, demonstrating the inherent difficulty of simultaneously controlling both type I and type II errors without additional assumptions. Nevertheless, under a monotone likelihood ratio (MLR) assumption, our auditing mechanism effectively controls both errors. We also extend our method to construct valid confidence bands for the trade-off function in the finite-sample regime.
Authors: Omri Isac, Idan Refaeli, Haoze Wu, Clark Barrett, Guy Katz
Abstract: The widespread adoption of deep neural networks (DNNs) requires efficient techniques for safety verification. Existing methods struggle to scale to real-world DNNs, and tremendous efforts are being put into improving their scalability. In this work, we propose an approach for improving the scalability of DNN verifiers using Conflict-Driven Clause Learning (CDCL) -- an approach that has proven highly successful in SAT and SMT solving. We present a novel algorithm for deriving conflict clauses using UNSAT proofs, and propose several optimizations for expediting it. Our approach allows a modular integration of SAT solvers and DNN verifiers, and we implement it on top of an interface designed for this purpose. The evaluation of our implementation over several benchmarks suggests a 2X--3X improvement over a similar approach, with specific cases outperforming the state of the art.
Authors: Matteo Cercola, Nicola Gatti, Pedro Huertas Leyva, Benedetto Carambia, Simone Formentin
Abstract: Effective traffic incident management is essential for ensuring safety, minimizing congestion, and reducing response times in emergency situations. Traditional highway incident management relies heavily on radio room operators, who must make rapid, informed decisions in high-stakes environments. This paper proposes an innovative solution to support and enhance these decisions by integrating Large Language Models (LLMs) into a decision-support system for traffic incident management. We introduce two approaches: (1) an LLM + Optimization hybrid that leverages both the flexibility of natural language interaction and the robustness of optimization techniques, and (2) a Full LLM approach that autonomously generates decisions using only LLM capabilities. We tested our solutions using historical event data from Autostrade per l'Italia. Experimental results indicate that while both approaches show promise, the LLM + Optimization solution demonstrates superior reliability, making it particularly suited to critical applications where consistency and accuracy are paramount. This research highlights the potential for LLMs to transform highway incident management by enabling accessible, data-driven decision-making support.
Authors: Shentong Mo, Zehua Chen, Fan Bao, Jun Zhu
Abstract: Recent works in cross-modal understanding and generation, notably through models like CLAP (Contrastive Language-Audio Pretraining) and CAVP (Contrastive Audio-Visual Pretraining), have significantly enhanced the alignment of text, video, and audio embeddings via a single contrastive loss. However, these methods often overlook the bidirectional interactions and inherent noises present in each modality, which can crucially impact the quality and efficacy of cross-modal integration. To address this limitation, we introduce DiffGAP, a novel approach incorporating a lightweight generative module within the contrastive space. Specifically, our DiffGAP employs a bidirectional diffusion process tailored to bridge the cross-modal gap more effectively. This involves a denoising process on text and video embeddings conditioned on audio embeddings and vice versa, thus facilitating a more nuanced and robust cross-modal interaction. Our experimental results on VGGSound and AudioCaps datasets demonstrate that DiffGAP significantly improves performance in video/text-audio generation and retrieval tasks, confirming its effectiveness in enhancing cross-modal understanding and generation capabilities.
Authors: Maryam Daniali, Shivaram Karandikar, Dabriel Zimmerman, J. Eric Schmitt, Matthew J. Buczek, Benjamin Jung, Laura Mercedes, Jakob Seidlitz, Vanessa Troiani, Lena Dorfschmidt, Eren Kafadar, Remo Williams, Susan Sotardi, Arastoo Vosough, Scott Haag, Jenna M. Schabdach, Aaron Alexander-Bloch
Abstract: Clinically acquired brain MRIs and radiology reports are valuable but underutilized resources due to the challenges of manual analysis and data heterogeneity. We developed fine-tuned language models (LMs) to classify brain MRI reports as normal (reports with limited pathology) or abnormal, fine-tuning BERT, BioBERT, ClinicalBERT, and RadBERT on 44,661 reports. We also explored the reasoning capabilities of a leading LM, Gemini 1.5-Pro, for normal report categorization. Automated image processing and modeling generated brain growth charts from LM-classified normal scans, comparing them to human-derived charts. Fine-tuned LMs achieved high classification performance (F1-Score >97%), with unbalanced training mitigating class imbalance. Performance was robust on out-of-distribution data, with full text outperforming summary (impression) sections. Gemini 1.5-Pro showed a promising categorization performance, especially with clinical inference. LM-derived brain growth charts were nearly identical to human-annotated charts (r = 0.99, p < 2.2e-16). Our LMs offer scalable analysis of radiology reports, enabling automated classification of brain MRIs in large datasets. One application is automated generation of brain growth charts for benchmarking quantitative image features. Further research is needed to address data heterogeneity and optimize LM reasoning.
Authors: Harold Triedman, Rishi Jha, Vitaly Shmatikov
Abstract: Multi-agent systems coordinate LLM-based agents to perform tasks on users' behalf. In real-world applications, multi-agent systems will inevitably interact with untrusted inputs, such as malicious Web content, files, email attachments, etc. Using several recently proposed multi-agent frameworks as concrete examples, we demonstrate that adversarial content can hijack control and communication within the system to invoke unsafe agents and functionalities. This results in a complete security breach, up to execution of arbitrary malicious code on the user's device and/or exfiltration of sensitive data from the user's containerized environment. We show that control-flow hijacking attacks succeed even if the individual agents are not susceptible to direct or indirect prompt injection, and even if they refuse to perform harmful actions.
Authors: S Balasubramanian, Yedu Krishna P, Talasu Sai Sriram, M Sai Subramaniam, Manepalli Pranav Phanindra Sai, Darshan Gera
Abstract: Feature Distillation (FD) strategies are proven to be effective in mitigating Catastrophic Forgetting (CF) seen in Class Incremental Learning (CIL). However, current FD approaches enforce strict alignment of feature magnitudes and directions across incremental steps, limiting the model's ability to adapt to new knowledge. In this paper we propose Structurally Stable Incremental Learning(S22IL), a FD method for CIL that mitigates CF by focusing on preserving the overall spatial patterns of features which promote flexible (plasticity) yet stable representations that preserve old knowledge (stability). We also demonstrate that our proposed method S2IL achieves strong incremental accuracy and outperforms other FD methods on SOTA benchmark datasets CIFAR-100, ImageNet-100 and ImageNet-1K. Notably, S2IL outperforms other methods by a significant margin in scenarios that have a large number of incremental tasks.
Authors: Mayank Kumar, Jiaqi Xue, Mengxin Zheng, Qian Lou
Abstract: Fully Homomorphic Encryption over the torus (TFHE) enables computation on encrypted data without decryption, making it a cornerstone of secure and confidential computing. Despite its potential in privacy preserving machine learning, secure multi party computation, private blockchain transactions, and secure medical diagnostics, its adoption remains limited due to cryptographic complexity and usability challenges. While various TFHE libraries and compilers exist, practical code generation remains a hurdle. We propose a compiler integrated framework to evaluate LLM inference and agentic optimization for TFHE code generation, focusing on logic gates and ReLU activation. Our methodology assesses error rates, compilability, and structural similarity across open and closedsource LLMs. Results highlight significant limitations in off-the-shelf models, while agentic optimizations such as retrieval augmented generation (RAG) and few-shot prompting reduce errors and enhance code fidelity. This work establishes the first benchmark for TFHE code generation, demonstrating how LLMs, when augmented with domain-specific feedback, can bridge the expertise gap in FHE code generation.
Authors: Daryna Chernobrovkina, Steffen Gr\"unew\"alder
Abstract: We analyze the prior that a Deep Gaussian Process with polynomial kernels induces. We observe that, even for relatively small depths, averaging effects occur within such a Deep Gaussian Process and that the prior can be analyzed and approximated effectively by means of the Berry-Esseen Theorem. One of the key findings of this analysis is that, in the absence of careful hyper-parameter tuning, the prior of a Deep Gaussian Process either collapses rapidly towards zero as the depth increases or places negligible mass on low norm functions. This aligns well with experimental findings and mirrors known results for convolution based Deep Gaussian Processes.
Authors: Siddharth Rout, Eldad Haber, St\'ephane Gaudreault
Abstract: The modeling of dynamical systems is essential in many fields, but applying machine learning techniques is often challenging due to incomplete or noisy data. This study introduces a variant of stochastic interpolation (SI) for probabilistic forecasting, estimating future states as distributions rather than single-point predictions. We explore its mathematical foundations and demonstrate its effectiveness on various dynamical systems, including the challenging WeatherBench dataset.
Authors: Paola Natalia Ca\~nas, Marcos Nieto, Oihana Otaegui, Igor Rodr\'iguez
Abstract: In recent years, we have witnessed significant progress in emerging deep learning models, particularly Large Language Models (LLMs) and Vision-Language Models (VLMs). These models have demonstrated promising results, indicating a new era of Artificial Intelligence (AI) that surpasses previous methodologies. Their extensive knowledge and zero-shot capabilities suggest a paradigm shift in developing deep learning solutions, moving from data capturing and algorithm training to just writing appropriate prompts. While the application of these technologies has been explored across various industries, including automotive, there is a notable gap in the scientific literature regarding their use in Driver Monitoring Systems (DMS). This paper presents our initial approach to implementing VLMs in this domain, utilising the Driver Monitoring Dataset to evaluate their performance and discussing their advantages and challenges when implemented in real-world scenarios.
Authors: Averi Bates, Ryan Vavricka, Shane Carleton, Ruosi Shao, Chongle Pan
Abstract: The Unified Modeling Language is a standardized visual language widely used for modeling and documenting the design of software systems. Although many tools generate UML diagrams from UML code, generating executable UML code from image-based UML diagrams remains challenging. This paper proposes a new approach to generate UML code using a large multimodal language model automatically. Synthetic UML activity and sequence diagram datasets were created to train and test the model. We compared standard fine-tuning with LoRA techniques to optimize base models. The experiments measured code generation accuracy across different model sizes and training strategies. These results demonstrated that domain-adapted MM-LLMs perform for UML code generation automation, whereby, at the best model, it achieved BLEU and SSIM scores of 0.779 and 0.942 on sequence diagrams. This will enable the modernization of legacy systems and decrease the manual effort in software development workflows.
Authors: Ryan P. Kelly, David J. Warne, David T. Frazier, David J. Nott, Michael U. Gutmann, Christopher Drovandi
Abstract: Simulation-based Bayesian inference (SBI) methods are widely used for parameter estimation in complex models where evaluating the likelihood is challenging but generating simulations is relatively straightforward. However, these methods commonly assume that the simulation model accurately reflects the true data-generating process, an assumption that is frequently violated in realistic scenarios. In this paper, we focus on the challenges faced by SBI methods under model misspecification. We consolidate recent research aimed at mitigating the effects of misspecification, highlighting three key strategies: i) robust summary statistics, ii) generalised Bayesian inference, and iii) error modelling and adjustment parameters. To illustrate both the vulnerabilities of popular SBI methods and the effectiveness of misspecification-robust alternatives, we present empirical results on an illustrative example.
Authors: Kumar Krishna Agrawal, Long Lian, Longchao Liu, Natalia Harguindeguy, Boyi Li, Alexander Bick, Maggie Chung, Trevor Darrell, Adam Yala
Abstract: Efficiently modeling massive images is a long-standing challenge in machine learning. To this end, we introduce Multi-Scale Attention (MSA). MSA relies on two key ideas, (i) multi-scale representations (ii) bi-directional cross-scale communication. MSA creates O(log N) scales to represent the image across progressively coarser features and leverages cross-attention to propagate information across scales. We then introduce Atlas, a novel neural network architecture based on MSA. We demonstrate that Atlas significantly improves the compute-performance tradeoff of long-context image modeling in a high-resolution variant of ImageNet 100. At 1024px resolution, Atlas-B achieves 91.04% accuracy, comparable to ConvNext-B (91.92%) while being 4.3x faster. Atlas is 2.95x faster and 7.38% better than FasterViT, 2.25x faster and 4.96% better than LongViT. In comparisons against MambaVision-S, we find Atlas-S achieves 5%, 16% and 32% higher accuracy at 1024px, 2048px and 4096px respectively, while obtaining similar runtimes. Code for reproducing our experiments and pretrained models is available at https://github.com/yalalab/atlas.
Authors: In-Chang Baek, Sung-Hyun Kim, Seo-yung Lee, Dong-Hyun Lee, Kyung-Joong Kim
Abstract: Recent research has highlighted the significance of natural language in enhancing the controllability of generative models. While various efforts have been made to leverage natural language for content generation, research on deep reinforcement learning (DRL) agents utilizing text-based instructions for procedural content generation remains limited. In this paper, we propose IPCGRL, an instruction-based procedural content generation method via reinforcement learning, which incorporates a sentence embedding model. IPCGRL fine-tunes task-specific embedding representations to effectively compress game-level conditions. We evaluate IPCGRL in a two-dimensional level generation task and compare its performance with a general-purpose embedding method. The results indicate that IPCGRL achieves up to a 21.4% improvement in controllability and a 17.2% improvement in generalizability for unseen instructions. Furthermore, the proposed method extends the modality of conditional input, enabling a more flexible and expressive interaction framework for procedural content generation.
Authors: Kuan-Lin Chen, Bhaskar D. Rao
Abstract: Covariance matrix reconstruction has been the most widely used guiding objective in gridless direction-of-arrival (DoA) estimation for sparse linear arrays. Many semidefinite programming (SDP)-based methods fall under this category. Although deep learning-based approaches enable the construction of more sophisticated objective functions, most methods still rely on covariance matrix reconstruction. In this paper, we propose new loss functions that are invariant to the scaling of the matrices and provide a comparative study of losses with varying degrees of invariance. The proposed loss functions are formulated based on the scale-invariant signal-to-distortion ratio between the target matrix and the Gram matrix of the prediction. Numerical results show that a scale-invariant loss outperforms its non-invariant counterpart but is inferior to the recently proposed subspace loss that is invariant to the change of basis. These results provide evidence that designing loss functions with greater degrees of invariance is advantageous in deep learning-based gridless DoA estimation.
Authors: Artem Lensky
Abstract: This study investigates EEG as a potential early biomarker by applying deep learning techniques to resting-state EEG recordings from 31 subjects (15 with PD and 16 healthy controls). EEG signals were rigorously preprocessed to remove tremor artifacts, then converted to wavelet-based images by grouping spatially adjacent electrodes into triplets for convolutional neural network (CNN) classification. Our analysis across different brain regions and frequency bands showed distinct spatial-spectral patterns of PD-related neural oscillations. We identified high classification accuracy (74%) in the gamma band (40-62.4 Hz) for central-parietal electrodes (CP1, Pz, CP2), and 76% accuracy using central electrodes (C3, Cz, C4) with full-spectrum 0.4-62.4 Hz. In particular, we observed pronounced right-hemisphere involvement, specifically in parieto-occipital regions. Unlike previous studies that achieved higher accuracies by potentially including tremor artifacts, our approach isolates genuine neurophysiological alterations in cortical activity. These findings suggest that specific EEG-based oscillatory patterns, especially central-parietal gamma activity, may provide diagnostic information for PD, potentially before the onset of motor symptoms.
Authors: Heng Zhang, Guoxiang Zhao, Xiaoqiang Ren
Abstract: Pursuit-evasion (PE) problem is a critical challenge in multi-robot systems (MRS). While reinforcement learning (RL) has shown its promise in addressing PE tasks, research has primarily focused on single-target pursuit, with limited exploration of multi-target encirclement, particularly in large-scale settings. This paper proposes a Transformer-Enhanced Reinforcement Learning (TERL) framework for large-scale multi-target encirclement. By integrating a transformer-based policy network with target selection, TERL enables robots to adaptively prioritize targets and safely coordinate robots. Results show that TERL outperforms existing RL-based methods in terms of encirclement success rate and task completion time, while maintaining good performance in large-scale scenarios. Notably, TERL, trained on small-scale scenarios (15 pursuers, 4 targets), generalizes effectively to large-scale settings (80 pursuers, 20 targets) without retraining, achieving a 100% success rate.
Authors: Martino Chiarani, Swastika Roy, Christos Verikoukis, Fabrizio Granelli
Abstract: In recent years, network slicing has embraced artificial intelligence (AI) models to manage the growing complexity of communication networks. In such a situation, AI-driven zero-touch network automation should present a high degree of flexibility and viability, especially when deployed in live production networks. However, centralized controllers suffer from high data communication overhead due to the vast amount of user data, and most network slices are reluctant to share private data. In federated learning systems, selecting trustworthy clients to participate in training is critical for ensuring system performance and reliability. The present paper proposes a new approach to client selection by leveraging an XAI method to guarantee scalable and fast operation of federated learning based analytic engines that implement slice-level resource provisioning at the RAN-Edge in a non-IID scenario. Attributions from XAI are used to guide the selection of devices participating in training. This approach enhances network trustworthiness for users and addresses the black-box nature of neural network models. The simulations conducted outperformed the standard approach in terms of both convergence time and computational cost, while also demonstrating high scalability.
Authors: Hossein Ranjbar, Alireza Taheri
Abstract: Sign language recognition involves modeling complex multichannel information, such as hand shapes and movements while relying on sufficient sign language-specific data. However, sign languages are often under-resourced, posing a significant challenge for research and development in this field. To address this gap, we introduce ISLR101, the first publicly available Iranian Sign Language dataset for isolated sign language recognition. This comprehensive dataset includes 4,614 videos covering 101 distinct signs, recorded by 10 different signers (3 deaf individuals, 2 sign language interpreters, and 5 L2 learners) against varied backgrounds, with a resolution of 800x600 pixels and a frame rate of 25 frames per second. It also includes skeleton pose information extracted using OpenPose. We establish both a visual appearance-based and a skeleton-based framework as baseline models, thoroughly training and evaluating them on ISLR101. These models achieve 97.01% and 94.02% accuracy on the test set, respectively. Additionally, we publish the train, validation, and test splits to facilitate fair comparisons.
Authors: Ke Chen, Dandan Jiang
Abstract: The process generates substantial amounts of data with highly complex structures, leading to the development of numerous nonlinear statistical methods. However, most of these methods rely on computations involving large-scale dense kernel matrices. This dependence poses significant challenges in meeting the high computational demands and real-time responsiveness required by online monitoring systems. To alleviate the computational burden of dense large-scale matrix multiplication, we incorporate the bootstrap sampling concept into random feature mapping and propose a novel random Bernoulli principal component analysis method to efficiently capture nonlinear patterns in the process. We derive a convergence bound for the kernel matrix approximation constructed using random Bernoulli features, ensuring theoretical robustness. Subsequently, we design four fast process monitoring methods based on random Bernoulli principal component analysis to extend its nonlinear capabilities for handling diverse fault scenarios. Finally, numerical experiments and real-world data analyses are conducted to evaluate the performance of the proposed methods. Results demonstrate that the proposed methods offer excellent scalability and reduced computational complexity, achieving substantial cost savings with minimal performance loss compared to traditional kernel-based approaches.
Authors: Alessio Xompero (Queen Mary University of London), Andrea Cavallaro (Idiap Research Institute, \'Ecole Polytechnique F\'ed\'erale de Lausanne)
Abstract: Subjective interpretation and content diversity make predicting whether an image is private or public a challenging task. Graph neural networks combined with convolutional neural networks (CNNs), which consist of 14,000 to 500 millions parameters, generate features for visual entities (e.g., scene and object types) and identify the entities that contribute to the decision. In this paper, we show that using a simpler combination of transfer learning and a CNN to relate privacy with scene types optimises only 732 parameters while achieving comparable performance to that of graph-based methods. On the contrary, end-to-end training of graph-based methods can mask the contribution of individual components to the classification performance. Furthermore, we show that a high-dimensional feature vector, extracted with CNNs for each visual entity, is unnecessary and complexifies the model. The graph component has also negligible impact on performance, which is driven by fine-tuning the CNN to optimise image features for privacy nodes.
Authors: Sebastian Reich
Abstract: We consider the problem of optimal control for partially observed dynamical systems. Despite its prevalence in practical applications, there are still very few algorithms available, which take uncertainties in the current state estimates and future observations into account. In other words, most current approaches separate state estimation from the optimal control problem. In this paper, we extend the popular ensemble Kalman filter to receding horizon optimal control problems in the spirit of nonlinear model predictive control. We provide an interacting particle approximation to the forward-backward stochastic differential equations arising from Pontryagin's maximum principle with the forward stochastic differential equation provided by the time-continuous ensemble Kalman-Bucy filter equations. The receding horizon control laws are approximated as linear and are continuously updated as in nonlinear model predictive control. We illustrate the performance of the proposed methodology for an inverted pendulum example.
Authors: Jiakang Chen, Selim F. Yilmaz, Di You, Pier Luigi Dragotti, Deniz G\"und\"uz
Abstract: Joint source-channel coding systems based on deep neural networks (DeepJSCC) have recently demonstrated remarkable performance in wireless image transmission. Existing methods primarily focus on minimizing distortion between the transmitted image and the reconstructed version at the receiver, often overlooking perceptual quality. This can lead to severe perceptual degradation when transmitting images under extreme conditions, such as low bandwidth compression ratios (BCRs) and low signal-to-noise ratios (SNRs). In this work, we propose SING, a novel two-stage JSCC framework that formulates the recovery of high-quality source images from corrupted reconstructions as an inverse problem. Depending on the availability of information about the DeepJSCC encoder/decoder and the channel at the receiver, SING can either approximate the stochastic degradation as a linear transformation, or leverage invertible neural networks (INNs) for precise modeling. Both approaches enable the seamless integration of diffusion models into the reconstruction process, enhancing perceptual quality. Experimental results demonstrate that SING outperforms DeepJSCC and other approaches, delivering superior perceptual quality even under extremely challenging conditions, including scenarios with significant distribution mismatches between the training and test data.
Authors: Pan Du, Delin An, Chaoli Wang, Jian-Xun Wang
Abstract: Image-based modeling is essential for understanding cardiovascular hemodynamics and advancing the diagnosis and treatment of cardiovascular diseases. Constructing patient-specific vascular models remains labor-intensive, error-prone, and time-consuming, limiting their clinical applications. This study introduces a deep-learning framework that automates the creation of simulation-ready vascular models from medical images. The framework integrates a segmentation module for accurate voxel-based vessel delineation with a surface deformation module that performs anatomically consistent and unsupervised surface refinements guided by medical image data. By unifying voxel segmentation and surface deformation into a single cohesive pipeline, the framework addresses key limitations of existing methods, enhancing geometric accuracy and computational efficiency. Evaluated on publicly available datasets, the proposed approach demonstrates state-of-the-art performance in segmentation and mesh quality while significantly reducing manual effort and processing time. This work advances the scalability and reliability of image-based computational modeling, facilitating broader applications in clinical and research settings.
Authors: Haoqi Yuan, Yu Bai, Yuhui Fu, Bohan Zhou, Yicheng Feng, Xinrun Xu, Yi Zhan, B\"orje F. Karlsson, Zongqing Lu
Abstract: Building autonomous robotic agents capable of achieving human-level performance in real-world embodied tasks is an ultimate goal in humanoid robot research. Recent advances have made significant progress in high-level cognition with Foundation Models (FMs) and low-level skill development for humanoid robots. However, directly combining these components often results in poor robustness and efficiency due to compounding errors in long-horizon tasks and the varied latency of different modules. We introduce Being-0, a hierarchical agent framework that integrates an FM with a modular skill library. The FM handles high-level cognitive tasks such as instruction understanding, task planning, and reasoning, while the skill library provides stable locomotion and dexterous manipulation for low-level control. To bridge the gap between these levels, we propose a novel Connector module, powered by a lightweight vision-language model (VLM). The Connector enhances the FM's embodied capabilities by translating language-based plans into actionable skill commands and dynamically coordinating locomotion and manipulation to improve task success. With all components, except the FM, deployable on low-cost onboard computation devices, Being-0 achieves efficient, real-time performance on a full-sized humanoid robot equipped with dexterous hands and active vision. Extensive experiments in large indoor environments demonstrate Being-0's effectiveness in solving complex, long-horizon tasks that require challenging navigation and manipulation subtasks. For further details and videos, visit https://beingbeyond.github.io/being-0.
Authors: Wei Zhu, Abirath Raju, Abdulaziz Shamsah, Anqi Wu, Seth Hutchinson, Ye Zhao
Abstract: This study presents an emotion-aware navigation framework -- EmoBipedNav -- using deep reinforcement learning (DRL) for bipedal robots walking in socially interactive environments. The inherent locomotion constraints of bipedal robots challenge their safe maneuvering capabilities in dynamic environments. When combined with the intricacies of social environments, including pedestrian interactions and social cues, such as emotions, these challenges become even more pronounced. To address these coupled problems, we propose a two-stage pipeline that considers both bipedal locomotion constraints and complex social environments. Specifically, social navigation scenarios are represented using sequential LiDAR grid maps (LGMs), from which we extract latent features, including collision regions, emotion-related discomfort zones, social interactions, and the spatio-temporal dynamics of evolving environments. The extracted features are directly mapped to the actions of reduced-order models (ROMs) through a DRL architecture. Furthermore, the proposed framework incorporates full-order dynamics and locomotion constraints during training, effectively accounting for tracking errors and restrictions of the locomotion controller while planning the trajectory with ROMs. Comprehensive experiments demonstrate that our approach exceeds both model-based planners and DRL-based baselines. The hardware videos and open-source code are available at https://gatech-lidar.github.io/emobipednav.github.io/.
URLs: https://gatech-lidar.github.io/emobipednav.github.io/.
Authors: Francesco Girlanda, Denys Rozumnyi, Marc Pollefeys, Martin R. Oswald
Abstract: We present Deblur-SLAM, a robust RGB SLAM pipeline designed to recover sharp reconstructions from motion-blurred inputs. The proposed method bridges the strengths of both frame-to-frame and frame-to-model approaches to model sub-frame camera trajectories that lead to high-fidelity reconstructions in motion-blurred settings. Moreover, our pipeline incorporates techniques such as online loop closure and global bundle adjustment to achieve a dense and precise global trajectory. We model the physical image formation process of motion-blurred images and minimize the error between the observed blurry images and rendered blurry images obtained by averaging sharp virtual sub-frame images. Additionally, by utilizing a monocular depth estimator alongside the online deformation of Gaussians, we ensure precise mapping and enhanced image deblurring. The proposed SLAM pipeline integrates all these components to improve the results. We achieve state-of-the-art results for sharp map estimation and sub-frame trajectory recovery both on synthetic and real-world blurry input data.
Authors: Emilio Cartoni, Gianluca Cioccolini, Gianluca Baldassarre
Abstract: Open-Ended Learning (OEL) autonomous robots can acquire new skills and knowledge through direct interaction with their environment, relying on mechanisms such as intrinsic motivations and self-generated goals to guide learning processes. OEL robots are highly relevant for applications as they can autonomously leverage acquired knowledge to perform tasks beneficial to human users in unstructured environments, addressing challenges unforeseen at design time. However, OEL robots face a significant limitation: their openness may lead them to waste time learning information that is irrelevant to tasks desired by specific users. Here, we propose a solution called `Purpose-Directed Open-Ended Learning' (POEL), based on the novel concept of `purpose' introduced in previous work. A purpose specifies what users want the robot to achieve. The key insight of this work is that purpose can focus OEL on learning self-generated classes of tasks that, while unknown during autonomous learning (as typical in OEL), involve objects relevant to the purpose. This concept is operationalised in a novel robot architecture capable of receiving a human purpose through speech-to-text, analysing the scene to identify objects, and using a Large Language Model to reason about which objects are purpose-relevant. These objects are then used to bias OEL exploration towards their spatial proximity and to self-generate rewards that favour interactions with them. The solution is tested in a simulated scenario where a camera-arm-gripper robot interacts freely with purpose-related and distractor objects. For the first time, the results demonstrate the potential advantages of purpose-focused OEL over state-of-the-art OEL methods, enabling robots to handle unstructured environments while steering their learning toward knowledge acquisition relevant to users.
Authors: Thayer Alshaabi, Daniel E. Milkie, Gaoxiang Liu, Cyna Shirazinejad, Jason L. Hong, Kemal Achour, Frederik G\"orlitz, Ana Milunovic-Jevtic, Cat Simmons, Ibrahim S. Abuzahriyeh, Erin Hong, Samara Erin Williams, Nathanael Harrison, Evan Huang, Eun Seok Bae, Alison N. Killilea, David G. Drubin, Ian A. Swinburne, Srigokul Upadhyayula, Eric Betzig
Abstract: High-resolution tissue imaging is often compromised by sample-induced optical aberrations that degrade resolution and contrast. While wavefront sensor-based adaptive optics (AO) can measure these aberrations, such hardware solutions are typically complex, expensive to implement, and slow when serially mapping spatially varying aberrations across large fields of view. Here, we introduce AOViFT (Adaptive Optical Vision Fourier Transformer) -- a machine learning-based aberration sensing framework built around a 3D multistage Vision Transformer that operates on Fourier domain embeddings. AOViFT infers aberrations and restores diffraction-limited performance in puncta-labeled specimens with substantially reduced computational cost, training time, and memory footprint compared to conventional architectures or real-space networks. We validated AOViFT on live gene-edited zebrafish embryos, demonstrating its ability to correct spatially varying aberrations using either a deformable mirror or post-acquisition deconvolution. By eliminating the need for the guide star and wavefront sensing hardware and simplifying the experimental workflow, AOViFT lowers technical barriers for high-resolution volumetric microscopy across diverse biological samples.
Authors: Alessio Spagnoletti, Jean Prost, Andr\'es Almansa, Nicolas Papadakis, Marcelo Pereyra
Abstract: Text-to-image latent diffusion models (LDMs) have recently emerged as powerful generative models with great potential for solving inverse problems in imaging. However, leveraging such models in a Plug & Play (PnP), zero-shot manner remains challenging because it requires identifying a suitable text prompt for the unknown image of interest. Also, existing text-to-image PnP approaches are highly computationally expensive. We herein address these challenges by proposing a novel PnP inference paradigm specifically designed for embedding generative models within stochastic inverse solvers, with special attention to Latent Consistency Models (LCMs), which distill LDMs into fast generators. We leverage our framework to propose LAtent consisTency INverse sOlver (LATINO), the first zero-shot PnP framework to solve inverse problems with priors encoded by LCMs. Our conditioning mechanism avoids automatic differentiation and reaches SOTA quality in as little as 8 neural function evaluations. As a result, LATINO delivers remarkably accurate solutions and is significantly more memory and computationally efficient than previous approaches. We then embed LATINO within an empirical Bayesian framework that automatically calibrates the text prompt from the observed measurements by marginal maximum likelihood estimation. Extensive experiments show that prompt self-calibration greatly improves estimation, allowing LATINO with PRompt Optimization to define new SOTAs in image reconstruction quality and computational efficiency.
Authors: Anthony Lamelas, Harrison Muchnic
Abstract: This study explores the impact of scaling semantic categories on the image classification performance of vision transformers (ViTs). In this specific case, the CLIP server provided by Jina AI is used for experimentation. The research hypothesizes that as the number of ground truth and artificially introduced semantically equivalent categories increases, the labeling accuracy of ViTs improves until a theoretical maximum or limit is reached. A wide variety of image datasets were chosen to test this hypothesis. These datasets were processed through a custom function in Python designed to evaluate the model's accuracy, with adjustments being made to account for format differences between datasets. By exponentially introducing new redundant categories, the experiment assessed accuracy trends until they plateaued, decreased, or fluctuated inconsistently. The findings show that while semantic scaling initially increases model performance, the benefits diminish or reverse after surpassing a critical threshold, providing insight into the limitations and possible optimization of category labeling strategies for ViTs.
Authors: Mohammad Al-Jarrah, Bamdad Hosseini, Amirhossein Taghvaei
Abstract: In this paper, we present the amortized optimal transport filter (A-OTF) designed to mitigate the computational burden associated with the real-time training of optimal transport filters (OTFs). OTFs can perform accurate non-Gaussian Bayesian updates in the filtering procedure, but they require training at every time step, which makes them expensive. The proposed A-OTF framework exploits the similarity between OTF maps during an initial/offline training stage in order to reduce the cost of inference during online calculations. More precisely, we use clustering algorithms to select relevant subsets of pre-trained maps whose weighted average is used to compute the A-OTF model akin to a mixture of experts. A series of numerical experiments validate that A-OTF achieves substantial computational savings during online inference while preserving the inherent flexibility and accuracy of OTF.
Authors: Imran Kabir, Md Alimoor Reza, Syed Billah
Abstract: Large multimodal models (LMMs) are increasingly integrated into autonomous driving systems for user interaction. However, their limitations in fine-grained spatial reasoning pose challenges for system interpretability and user trust. We introduce Logic-RAG, a novel Retrieval-Augmented Generation (RAG) framework that improves LMMs' spatial understanding in driving scenarios. Logic-RAG constructs a dynamic knowledge base (KB) about object-object relationships in first-order logic (FOL) using a perception module, a query-to-logic embedder, and a logical inference engine. We evaluated Logic-RAG on visual-spatial queries using both synthetic and real-world driving videos. When using popular LMMs (GPT-4V, Claude 3.5) as proxies for an autonomous driving system, these models achieved only 55% accuracy on synthetic driving scenes and under 75% on real-world driving scenes. Augmenting them with Logic-RAG increased their accuracies to over 80% and 90%, respectively. An ablation study showed that even without logical inference, the fact-based context constructed by Logic-RAG alone improved accuracy by 15%. Logic-RAG is extensible: it allows seamless replacement of individual components with improved versions and enables domain experts to compose new knowledge in both FOL and natural language. In sum, Logic-RAG addresses critical spatial reasoning deficiencies in LMMs for autonomous driving applications. Code and data are available at https://github.com/Imran2205/LogicRAG.
Authors: Jeremy A. McCulloch, Ellen Kuhl
Abstract: When characterizing materials, it can be important to not only predict their mechanical properties, but also to estimate the probability distribution of these properties across a set of samples. Constitutive neural networks allow for the automated discovery of constitutive models that exactly satisfy physical laws given experimental testing data, but are only capable of predicting the mean stress response. Stochastic methods treat each weight as a random variable and are capable of learning their probability distributions. Bayesian constitutive neural networks combine both methods, but their weights lack physical interpretability and we must sample each weight from a probability distribution to train or evaluate the model. Here we introduce a more interpretable network with fewer parameters, simpler training, and the potential to discover correlated weights: Gaussian constitutive neural networks. We demonstrate the performance of our new Gaussian network on biaxial testing data, and discover a sparse and interpretable four-term model with correlated weights. Importantly, the discovered distributions of material parameters across a set of samples can serve as priors to discover better constitutive models for new samples with limited data. We anticipate that Gaussian constitutive neural networks are a natural first step towards generative constitutive models informed by physical laws and parameter uncertainty.
Authors: Tianyuan Wang
Abstract: In industrial X-ray Computed Tomography (CT), the need for rapid in-line inspection is critical. Sparse-angle tomography plays a significant role in this by reducing the required number of projections, thereby accelerating processing and conserving resources. Most existing methods aim to balance reconstruction quality and scanning time, typically relying on fixed scan durations. Adaptive adjustment of the number of angles is essential; for instance, more angles may be required for objects with complex geometries or noisier projections. The concept of optimal stopping, which dynamically adjusts this balance according to varying industrial needs, remains underutilized. Building on our previous work, we integrate optimal stopping into sequential Optimal Experimental Design (OED). We propose a novel method for computing the policy gradient within the Actor-Critic framework, enabling the development of adaptive policies for informative angle selection and scan termination. Additionally, we investigated the gap between simulation and real-world applications in the context of the developed learning-based method. Our trained model, developed using synthetic data, demonstrates reliable performance when applied to real-world data. This approach enhances the flexibility of CT operations and expands the applicability of sparse-angle tomography in industrial settings.
Authors: Jeihee Cho, Junyong Lee, Daniel Justice, Shiho Kim
Abstract: Hybrid quantum-classical computing relies heavily on Variational Quantum Algorithms (VQAs) to tackle challenges in diverse fields like quantum chemistry and machine learning. However, VQAs face a critical limitation: the balance between circuit trainability and expressibility. Trainability, the ease of optimizing circuit parameters for problem-solving, is often hampered by the Barren Plateau, where gradients vanish and hinder optimization. On the other hand, increasing expressibility, the ability to represent a wide range of quantum states, often necessitates deeper circuits with more parameters, which in turn exacerbates trainability issues. In this work, we investigate selective gate activation strategies as a potential solution to these challenges within the context of Variational Quantum Eigensolvers (VQEs). We evaluate three different approaches: activating gates randomly without considering their type or parameter magnitude, activating gates randomly but limited to a single gate type, and activating gates based on the magnitude of their parameter values. Experiment results reveal that the Magnitude-based strategy surpasses other methods, achieving improved convergence.
Authors: Ahmad M. Nagib, Hatem Abou-Zeid, Hossam S. Hassanein
Abstract: Deep reinforcement learning (DRL)-based slicing policies have shown significant success in simulated environments but face challenges in physical systems such as open radio access networks (O-RANs) due to simulation-to-reality gaps. These policies often lack safety guarantees to ensure compliance with service level agreements (SLAs), such as the strict latency requirements of immersive applications. As a result, a deployed DRL slicing agent may make resource allocation (RA) decisions that degrade system performance, particularly in previously unseen scenarios. Real-world immersive applications require maintaining SLA constraints throughout deployment to prevent risky DRL exploration. In this paper, we propose SafeSlice to address both the cumulative (trajectory-wise) and instantaneous (state-wise) latency constraints of O-RAN slices. We incorporate the cumulative constraints by designing a sigmoid-based risk-sensitive reward function that reflects the slices' latency requirements. Moreover, we build a supervised learning cost model as part of a safety layer that projects the slicing agent's RA actions to the nearest safe actions, fulfilling instantaneous constraints. We conduct an exhaustive experiment that supports multiple services, including real virtual reality (VR) gaming traffic, to investigate the performance of SafeSlice under extreme and changing deployment conditions. SafeSlice achieves reductions of up to 83.23% in average cumulative latency, 93.24% in instantaneous latency violations, and 22.13% in resource consumption compared to the baselines. The results also indicate SafeSlice's robustness to changing the threshold configurations of latency constraints, a vital deployment scenario that will be realized by the O-RAN paradigm to empower mobile network operators (MNOs).
Authors: Brian Cho, Ana-Roxana Pop, Ariel Evince, Nathan Kallus
Abstract: To design effective digital interventions, experimenters face the challenge of learning decision policies that balance multiple objectives using offline data. Often, they aim to develop policies that maximize goal outcomes, while ensuring there are no undesirable changes in guardrail outcomes. To provide credible recommendations, experimenters must not only identify policies that satisfy the desired changes in goal and guardrail outcomes, but also offer probabilistic guarantees about the changes these policies induce. In practice, however, policy classes are often large, and digital experiments tend to produce datasets with small effect sizes relative to noise. In this setting, standard approaches such as data splitting or multiple testing often result in unstable policy selection and/or insufficient statistical power. In this paper, we provide safe noisy policy learning (SNPL), a novel approach that leverages the concept of algorithmic stability to address these challenges. Our method enables policy learning while simultaneously providing high-confidence guarantees using the entire dataset, avoiding the need for data-splitting. We present finite-sample and asymptotic versions of our algorithm that ensure the recommended policy satisfies high-probability guarantees for avoiding guardrail regressions and/or achieving goal outcome improvements. We test both variants of our approach approach empirically on a real-world application of personalizing SMS delivery. Our results on real-world data suggest that our approach offers dramatic improvements in settings with large policy classes and low signal-to-noise across both finite-sample and asymptotic safety guarantees, offering up to 300\% improvements in detection rates and 150\% improvements in policy gains at significantly smaller sample sizes.
Authors: Kewei Sui, Anindita Ghosh, Inwoo Hwang, Jian Wang, Chuan Guo
Abstract: Humans inhabit a world defined by interactions -- with other humans, objects, and environments. These interactive movements not only convey our relationships with our surroundings but also demonstrate how we perceive and communicate with the real world. Therefore, replicating these interaction behaviors in digital systems has emerged as an important topic for applications in robotics, virtual reality, and animation. While recent advances in deep generative models and new datasets have accelerated progress in this field, significant challenges remain in modeling the intricate human dynamics and their interactions with entities in the external world. In this survey, we present, for the first time, a comprehensive overview of the literature in human interaction motion generation. We begin by establishing foundational concepts essential for understanding the research background. We then systematically review existing solutions and datasets across three primary interaction tasks -- human-human, human-object, and human-scene interactions -- followed by evaluation metrics. Finally, we discuss open research directions and future opportunities.
Authors: Weiming Xu, Tao Yang, Chang Liu, Kun Wu, Peng Zhang
Abstract: The scramjet engine is a key propulsion system for hypersonic vehicles, leveraging supersonic airflow to achieve high specific impulse, making it a promising technology for aerospace applications. Understanding and controlling the complex interactions between fuel injection, turbulent combustion, and aerodynamic effects of compressible flows are crucial for ensuring stable combustion in scramjet engines. However, identifying stable modes in scramjet combustors is often challenging due to limited experimental measurement means and extremely complex spatiotemporal evolution of supersonic turbulent combustion. This work introduces an innovative deep learning framework that combines dimensionality reduction via the Residual Convolutional Neural Network-beta-Variational Autoencoder (Res-CNN-beta-VAE) model with unsupervised clustering (K-means) to identify and analyze dynamical combustion modes in a supersonic combustor. By mapping high-dimensional data of combustion snapshots to a reduced three-dimensional latent space, the Res-CNN-beta-VAE model captures the essential temporal and spatial features of flame behaviors and enables the observation of transitions between combustion states. By analyzing the standard deviation of latent variable trajectories, we introduce a novel method for objectively distinguishing between dynamic transitions, which provides a scalable and expert-independent alternative to traditional classification methods. Besides, the unsupervised K-means clustering approach effectively identifies the complex interplay between the cavity and the jet-wake stabilization mechanisms, offering new insights into the system's behavior across different gas-to-liquid mass flow ratios (GLRs).
Authors: Chang Liu, Bavesh Balaji, Saad Hossain, C Thomas, Kwei-Herng Lai, Raviteja Vemulapalli, Alexander Wong, Sirisha Rambhatla
Abstract: Unsupervised domain adaptation for semantic segmentation (DASS) aims to transfer knowledge from a label-rich source domain to a target domain with no labels. Two key approaches in DASS are (1) vision-only approaches using masking or multi-resolution crops, and (2) language-based approaches that use generic class-wise prompts informed by target domain (e.g. "a {snowy} photo of a {class}"). However, the former is susceptible to noisy pseudo-labels that are biased to the source domain. The latter does not fully capture the intricate spatial relationships of objects -- key for dense prediction tasks. To this end, we propose LangDA. LangDA addresses these challenges by, first, learning contextual relationships between objects via VLM-generated scene descriptions (e.g. "a pedestrian is on the sidewalk, and the street is lined with buildings."). Second, LangDA aligns the entire image features with text representation of this context-aware scene caption and learns generalized representations via text. With this, LangDA sets the new state-of-the-art across three DASS benchmarks, outperforming existing methods by 2.6%, 1.4% and 3.9%.
Authors: Jingzhou Huang, Jiuyao Lu, Alexander Williams Tolbert
Abstract: Variable selection poses a significant challenge in causal modeling, particularly within the social sciences, where constructs often rely on inter-related factors such as age, socioeconomic status, gender, and race. Indeed, it has been argued that such attributes must be modeled as macro-level abstractions of lower-level manipulable features, in order to preserve the modularity assumption essential to causal inference. This paper accordingly extends the theoretical framework of Causal Feature Learning (CFL). Empirically, we apply the CFL algorithm to diverse social science datasets, evaluating how CFL-derived macrostates compare with traditional microstates in downstream modeling tasks.
Authors: Chen Li, Debo Cheng, Yasuhiko Morimoto
Abstract: Aspect-based sentiment analysis seeks to determine sentiment with a high level of detail. While graph convolutional networks (GCNs) are commonly used for extracting sentiment features, their straightforward use in syntactic feature extraction can lead to a loss of crucial information. This paper presents a novel edge-enhanced GCN, called EEGCN, which improves performance by preserving feature integrity as it processes syntactic graphs. We incorporate a bidirectional long short-term memory (Bi-LSTM) network alongside a self-attention-based transformer for effective text encoding, ensuring the retention of long-range dependencies. A bidirectional GCN (Bi-GCN) with message passing then captures the relationships between entities, while an aspect-specific masking technique removes extraneous information. Extensive evaluations and ablation studies on four benchmark datasets show that EEGCN significantly enhances aspect-based sentiment analysis, overcoming issues with syntactic feature extraction and advancing the field's methodologies.
Authors: Milind Nakul, Vidya Muthukumar, Ashwin Pananjady
Abstract: Suppose we observe a trajectory of length $n$ from an $\alpha$-mixing stochastic process over a finite but potentially large state space. We consider the problem of estimating the probability mass placed by the stationary distribution of any such process on elements that occur with a certain frequency in the observed sequence. We estimate this vector of probabilities in total variation distance, showing universal consistency in $n$ and recovering known results for i.i.d. sequences as special cases. Our proposed methodology carefully combines the plug-in (or empirical) estimator with a recently-proposed modification of the Good--Turing estimator called \textsc{WingIt}, which was originally developed for Markovian sequences. En route to controlling the error of our estimator, we develop new performance bounds on \textsc{WingIt} and the plug-in estimator for $\alpha$-mixing stochastic processes. Importantly, the extensively used method of Poissonization can no longer be applied in our non i.i.d. setting, and so we develop complementary tools -- including concentration inequalities for a natural self-normalized statistic of mixing sequences -- that may prove independently useful in the design and analysis of estimators for related problems.
Authors: Mousa Alizadeh, Mohammad Hossein Samaei, Azam Seilsepour, Mohammad TH Beheshti
Abstract: Effective epidemic modeling is essential for managing public health crises, requiring robust methods to predict disease spread and optimize resource allocation. This study introduces a novel deep learning framework that advances time series forecasting for infectious diseases, with its application to COVID 19 data as a critical case study. Our hybrid approach integrates Convolutional Neural Networks (CNNs) and Long Short Term Memory (LSTM) models to capture spatial and temporal dynamics of disease transmission across diverse regions. The CNN extracts spatial features from raw epidemiological data, while the LSTM models temporal patterns, yielding precise and adaptable predictions. To maximize performance, we employ a hybrid optimization strategy combining the Whale Optimization Algorithm (WOA) and Gray Wolf Optimization (GWO) to fine tune hyperparameters, such as learning rates, batch sizes, and training epochs enhancing model efficiency and accuracy. Applied to COVID 19 case data from 24 countries across six continents, our method outperforms established benchmarks, including ARIMA and standalone LSTM models, with statistically significant gains in predictive accuracy (e.g., reduced RMSE). This framework demonstrates its potential as a versatile method for forecasting epidemic trends, offering insights for resource planning and decision making in both historical contexts, like the COVID 19 pandemic, and future outbreaks.
Authors: Yanlin Xiang, Qingyuan He, Ting Xu, Ran Hao, Jiacheng Hu, Hanchao Zhang
Abstract: This study proposes a 3D semantic segmentation method for the spine based on the improved SwinUNETR to improve segmentation accuracy and robustness. Aiming at the complex anatomical structure of spinal images, this paper introduces a multi-scale fusion mechanism to enhance the feature extraction capability by using information of different scales, thereby improving the recognition accuracy of the model for the target area. In addition, the introduction of the adaptive attention mechanism enables the model to dynamically adjust the attention to the key area, thereby optimizing the boundary segmentation effect. The experimental results show that compared with 3D CNN, 3D U-Net, and 3D U-Net + Transformer, the model of this study has achieved significant improvements in mIoU, mDice, and mAcc indicators, and has better segmentation performance. The ablation experiment further verifies the effectiveness of the proposed improved method, proving that multi-scale fusion and adaptive attention mechanism have a positive effect on the segmentation task. Through the visualization analysis of the inference results, the model can better restore the real anatomical structure of the spinal image. Future research can further optimize the Transformer structure and expand the data scale to improve the generalization ability of the model. This study provides an efficient solution for the task of medical image segmentation, which is of great significance to intelligent medical image analysis.
Authors: Duke Nguyen, Aditya Joshi, Flora Salim
Abstract: Test-time adaptation (TTA) is an excellent method which helps generalize models across domains, tasks, and distributions without the use of labeled datasets. Thus, TTA is very useful in natural language processing (NLP) in the dialectal setting, since oftentimes, models are trained on Standard American English (SAE), evaluated on Indian English or Nigerian English, of which distribution differs significantly from the former. This is especially useful since dialectal datasets are scarce. In this paper, we explore one of the most famous TTA techniques, SHOT, in dialectal NLP. We finetune and evaluate SHOT on different combinations of dialectal GLUE. Our findings show that SHOT is a viable technique when labeled datasets are unavailable. We also theoretically propose the concept of dialectal gap and show that it has a positive correlation with the effectiveness of SHOT. We also find that in many cases, finetuning on SAE yields higher performance than finetuning on dialectal data. Our code is available at https://github.com/dukenguyenxyz/dialect-adaptation
Authors: Cho Hyeonsu, Dooyoung Kim, Youngjoong Ko
Abstract: There have been attempts to utilize linear probe for detoxification, with existing studies relying on a single toxicity probe vector to reduce toxicity. However, toxicity can be fine-grained into various subcategories, making it difficult to remove certain types of toxicity by using a single toxicity probe vector. To address this limitation, we propose a category-specific toxicity probe vector approach. First, we train multiple toxicity probe vectors for different toxicity categories. During generation, we dynamically select the most relevant toxicity probe vector based on the current context. Finally, the selected vector is dynamically scaled and subtracted from model. Our method successfully mitigated toxicity from categories that the single probe vector approach failed to detoxify. Experiments demonstrate that our approach achieves up to a 78.52% reduction in toxicity on the evaluation dataset, while fluency remains nearly unchanged, with only a 0.052% drop compared to the unsteered model.
Authors: Masanari Kimura
Abstract: We develop a higher-order asymptotic analysis for the semi-hard triplet loss using the Edgeworth expansion. It is known that this loss function enforces that embeddings of similar samples are close while those of dissimilar samples are separated by a specified margin. By refining the classical central limit theorem, our approach quantifies the impact of the margin parameter and the skewness of the underlying data distribution on the loss behavior. In particular, we derive explicit Edgeworth expansions that reveal first-order corrections in terms of the third cumulant, thereby characterizing non-Gaussian effects present in the distribution of distance differences between anchor-positive and anchor-negative pairs. Our findings provide detailed insight into the sensitivity of the semi-hard triplet loss to its parameters and offer guidance for choosing the margin to ensure training stability.
Authors: Jian Gu, Aldeida Aleti, Chunyang Chen, Hongyu Zhang
Abstract: Language Models (LMs) are widely used in software engineering for code generation, but they may produce code with errors. Rather than repairing the generated code, an alternative way is to address the underlying failures of models. LM repair offers a lightweight solution to this challenge: it requires minimal data, reduces computational costs, and reduces the side effects. Unlike retraining, LM repair focuses on applying tailored updates to targeted neurons, making it ideal for scenarios with limited resources, high-performance demands, or strict safety requirements. In this paper, we propose \ul{S}emantic \ul{T}argeting for \ul{A}nalytical \ul{R}epair (\textsc{STAR}), a pioneering and novel semantic-based optimization approach for repairing LLMs. \textsc{STAR} realizes main operations in LM repair methods in an optimization process, including locating ``buggy neurons'', solving ``neuron patches'', and patching ``buggy neurons''. Correspondingly, it computes the deltas of weight matrix as the prior information to guide optimization; and attributes the targeted layers and neurons leveraging statistical insights. The neuron patches are computed with a solid semantic-based analytical formula, which directly bridges the changes to logits with the deltas of neurons, by steering latent representations. Compared to the prior work of LM repair (\textsc{MINT}) and optimization methods (\textsc{SGD}), \textsc{STAR} integrates their strengths while mitigating their limitations. \textsc{STAR} supports solving multiple failures together, significantly improving the usefulness. Evaluated on three code generation tasks using popular code LMs, \textsc{STAR} demonstrates superior effectiveness. Additionally, \textsc{STAR} exhibits better efficiency. In terms of side effects, namely the balance between generalization and specificity, \textsc{STAR} outperforms prior work by a significant margin.
Authors: Bin Tang, Keqi Pan, Miao Zheng, Ning Zhou, Jialu Sui, Dandan Zhu, Cheng-Long Deng, Shu-Guang Kuai
Abstract: In recent years, predicting Big Five personality traits from multimodal data has received significant attention in artificial intelligence (AI). However, existing computational models often fail to achieve satisfactory performance. Psychological research has shown a strong correlation between pose and personality traits, yet previous research has largely ignored pose data in computational models. To address this gap, we develop a novel multimodal dataset that incorporates full-body pose data. The dataset includes video recordings of 287 participants completing a virtual interview with 36 questions, along with self-reported Big Five personality scores as labels. To effectively utilize this multimodal data, we introduce the Psychology-Inspired Network (PINet), which consists of three key modules: Multimodal Feature Awareness (MFA), Multimodal Feature Interaction (MFI), and Psychology-Informed Modality Correlation Loss (PIMC Loss). The MFA module leverages the Vision Mamba Block to capture comprehensive visual features related to personality, while the MFI module efficiently fuses the multimodal features. The PIMC Loss, grounded in psychological theory, guides the model to emphasize different modalities for different personality dimensions. Experimental results show that the PINet outperforms several state-of-the-art baseline models. Furthermore, the three modules of PINet contribute almost equally to the model's overall performance. Incorporating pose data significantly enhances the model's performance, with the pose modality ranking mid-level in importance among the five modalities. These findings address the existing gap in personality-related datasets that lack full-body pose data and provide a new approach for improving the accuracy of personality prediction models, highlighting the importance of integrating psychological insights into AI frameworks.
Authors: Jingyi Zhang, Jiaxing Huang, Huanjin Yao, Shunyu Liu, Xikun Zhang, Shijian Lu, Dacheng Tao
Abstract: Recent studies generally enhance MLLMs' reasoning capabilities via supervised fine-tuning on high-quality chain-of-thought reasoning data, which often leads models to merely imitate successful reasoning paths without understanding what the wrong reasoning paths are. In this work, we aim to enhance the MLLMs' reasoning ability beyond passively imitating positive reasoning paths. To this end, we design Step-wise Group Relative Policy Optimization (StepGRPO), a new online reinforcement learning framework that enables MLLMs to self-improve reasoning ability via simple, effective and dense step-wise rewarding. Specifically, StepGRPO introduces two novel rule-based reasoning rewards: Step-wise Reasoning Accuracy Reward (StepRAR) and Step-wise Reasoning Validity Reward (StepRVR). StepRAR rewards the reasoning paths that contain necessary intermediate reasoning steps via a soft key-step matching technique, while StepRAR rewards reasoning paths that follow a well-structured and logically consistent reasoning process through a reasoning completeness and logic evaluation strategy. With the proposed StepGRPO, we introduce R1-VL, a series of MLLMs with outstanding capabilities in step-by-step reasoning. Extensive experiments over 8 benchmarks demonstrate the superiority of our methods.
Authors: Haiyang Guo, Fanhu Zeng, Ziwei Xiang, Fei Zhu, Da-Han Wang, Xu-Yao Zhang, Cheng-Lin Liu
Abstract: Instruction tuning is widely used to improve a pre-trained Multimodal Large Language Model (MLLM) by training it on curated task-specific datasets, enabling better comprehension of human instructions. However, it is infeasible to collect all possible instruction datasets simultaneously in real-world scenarios. Thus, enabling MLLM with continual instruction tuning is essential for maintaining their adaptability. However, existing methods often trade off memory efficiency for performance gains, significantly compromising overall efficiency. In this paper, we propose a task-specific expansion and task-general fusion framework based on the variations in Centered Kernel Alignment (CKA) similarity across different model layers when trained on diverse datasets. Furthermore, we analyze the information leakage present in the existing benchmark and propose a new and more challenging benchmark to rationally evaluate the performance of different methods. Comprehensive experiments showcase a significant performance improvement of our method compared to existing state-of-the-art methods. Our code will be public available.
Authors: Zeeshan Patel, Ethan He, Parth Mannan, Xiaowei Ren, Ryan Wolf, Niket Agarwal, Jacob Huffman, Zhuoyao Wang, Carl Wang, Jack Chang, Yan Bai, Tommy Huang, Linnan Wang, Sahil Jain, Shanmugam Ramasamy, Joseph Jennings, Ekaterina Sirazitdinova, Oleg Sudakov, Mingyuan Ma, Bobby Chen, Forrest Lin, Hao Wang, Vasanth Rao Naik Sabavat, Sriharsha Niverty, Rong Ou, Pallab Bhattacharya, David Page, Nima Tajbakhsh, Ashwath Aithal
Abstract: Video Foundation Models (VFMs) have recently been used to simulate the real world to train physical AI systems and develop creative visual experiences. However, there are significant challenges in training large-scale, high quality VFMs that can generate high-quality videos. We present a scalable, open-source VFM training pipeline with NVIDIA NeMo, providing accelerated video dataset curation, multimodal data loading, and parallelized video diffusion model training and inference. We also provide a comprehensive performance analysis highlighting best practices for efficient VFM training and inference.
Authors: Muhan Hou, Koen Hindriks, A. E. Eiben, Kim Baraka
Abstract: Transfer Learning (TL) is a powerful tool that enables robots to transfer learned policies across different environments, tasks, or embodiments. To further facilitate this process, efforts have been made to combine it with Learning from Demonstrations (LfD) for more flexible and efficient policy transfer. However, these approaches are almost exclusively limited to offline demonstrations collected before policy transfer starts, which may suffer from the intrinsic issue of covariance shift brought by LfD and harm the performance of policy transfer. Meanwhile, extensive work in the learning-from-scratch setting has shown that online demonstrations can effectively alleviate covariance shift and lead to better policy performance with improved sample efficiency. This work combines these insights to introduce online demonstrations into a policy transfer setting. We present Policy Transfer with Online Demonstrations, an active LfD algorithm for policy transfer that can optimize the timing and content of queries for online episodic expert demonstrations under a limited demonstration budget. We evaluate our method in eight robotic scenarios, involving policy transfer across diverse environment characteristics, task objectives, and robotic embodiments, with the aim to transfer a trained policy from a source task to a related but different target task. The results show that our method significantly outperforms all baselines in terms of average success rate and sample efficiency, compared to two canonical LfD methods with offline demonstrations and one active LfD method with online demonstrations. Additionally, we conduct preliminary sim-to-real tests of the transferred policy on three transfer scenarios in the real-world environment, demonstrating the policy effectiveness on a real robot manipulator.
Authors: Etienne Gauthier, Francis Bach, Michael I. Jordan
Abstract: Conformal prediction is a powerful framework for distribution-free uncertainty quantification. The standard approach to conformal prediction relies on comparing the ranks of prediction scores: under exchangeability, the rank of a future test point cannot be too extreme relative to a calibration set. This rank-based method can be reformulated in terms of p-values. In this paper, we explore an alternative approach based on e-values, known as conformal e-prediction. E-values offer key advantages that cannot be achieved with p-values, enabling new theoretical and practical capabilities. In particular, we present three applications that leverage the unique strengths of e-values: batch anytime-valid conformal prediction, fixed-size conformal sets with data-dependent coverage, and conformal prediction under ambiguous ground truth. Overall, these examples demonstrate that e-value-based constructions provide a flexible expansion of the toolbox of conformal prediction.
Authors: Richard Biegler-K\"onig, Daniel Oeltz
Abstract: In power markets, Green Power Purchase Agreements have become an important contractual tool of the energy transition from fossil fuels to renewable sources such as wind or solar radiation. Trading Green PPAs exposes agents to price risks and weather risks. Also, developed electricity markets feature the so-called cannibalisation effect : large infeeds induce low prices and vice versa. As weather is a non-tradable entity the question arises how to hedge and risk-manage in this highly incom-plete setting. We propose a ''deep hedging'' framework utilising machine learning methods to construct hedging strategies. The resulting strategies outperform static and dynamic benchmark strategies with respect to different risk measures.
Authors: Erik Daxberger, Nina Wenzel, David Griffiths, Haiming Gang, Justin Lazarow, Gefen Kohavi, Kai Kang, Marcin Eichner, Yinfei Yang, Afshin Dehghan, Peter Grasch
Abstract: Multimodal large language models (MLLMs) excel at 2D visual understanding but remain limited in their ability to reason about 3D space. In this work, we leverage large-scale high-quality 3D scene data with open-set annotations to introduce 1) a novel supervised fine-tuning dataset and 2) a new evaluation benchmark, focused on indoor scenes. Our Cubify Anything VQA (CA-VQA) data covers diverse spatial tasks including spatial relationship prediction, metric size and distance estimation, and 3D grounding. We show that CA-VQA enables us to train MM-Spatial, a strong generalist MLLM that also achieves state-of-the-art performance on 3D spatial understanding benchmarks, including our own. We show how incorporating metric depth and multi-view inputs (provided in CA-VQA) can further improve 3D understanding, and demonstrate that data alone allows our model to achieve depth perception capabilities comparable to dedicated monocular depth estimation models. We will publish our SFT dataset and benchmark.
Authors: Zeng Wang, Minghao Shao, Mohammed Nabeel, Prithwish Basu Roy, Likhitha Mankali, Jitendra Bhandari, Ramesh Karri, Ozgur Sinanoglu, Muhammad Shafique, Johann Knechtel
Abstract: Large language models (LLMs) offer significant potential for coding, yet fine-tuning (FT) with curated data is essential for niche languages like Verilog. Using proprietary intellectual property (IP) for FT presents a serious risk, as FT data can be leaked through LLM inference. This leads to a critical dilemma for design houses: seeking to build externally accessible LLMs offering competitive Verilog coding, how can they leverage in-house IP to enhance FT utility while ensuring IP protection? For the first time in the literature, we study this dilemma. Using LLaMA 3.1-8B, we conduct in-house FT on a baseline Verilog dataset (RTLCoder) supplemented with our own in-house IP, which is validated through multiple tape-outs. To rigorously assess IP leakage, we quantify structural similarity (AST/Dolos) and functional equivalence (Synopsys Formality) between generated codes and our in-house IP. We show that our IP can indeed be leaked, confirming the threat. As defense, we evaluate logic locking of Verilog codes (ASSURE). This offers some level of protection, yet reduces the IP's utility for FT and degrades the LLM's performance. Our study shows the need for novel strategies that are both effective and minimally disruptive to FT, an essential effort for enabling design houses to fully utilize their proprietary IP toward LLM-driven Verilog coding.
Authors: Zhanggen Jin, Haobin Duan, Zhiyang Hang
Abstract: Games have played a pivotal role in advancing artificial intelligence, with AI agents using sophisticated techniques to compete. Despite the success of neural network based game AIs, their performance often requires significant computational resources. In this paper, we present Rapfi, an efficient Gomoku agent that outperforms CNN-based agents in limited computation environments. Rapfi leverages a compact neural network with a pattern-based codebook distilled from CNNs, and an incremental update scheme that minimizes computation when input changes are minor. This new network uses computation that is orders of magnitude less to reach a similar accuracy of much larger neural networks such as Resnet. Thanks to our incremental update scheme, depth-first search methods such as the alpha-beta search can be significantly accelerated. With a carefully tuned evaluation and search, Rapfi reached strength surpassing Katagomo, the strongest open-source Gomoku AI based on AlphaZero's algorithm, under limited computational resources where accelerators like GPUs are absent. Rapfi ranked first among 520 Gomoku agents on Botzone and won the championship in GomoCup 2024.
Authors: Corina Catarau-Cotutiu, Esther Mondragon, Eduardo Alonso
Abstract: The ability of artificial intelligence agents to make optimal decisions and generalise them to different domains and tasks is compromised in complex scenarios. One way to address this issue has focused on learning efficient representations of the world and on how the actions of agents affect them, such as disentangled representations that exploit symmetries. Whereas such representations are procedurally efficient, they are based on the compression of low-level state-action transitions, which lack structural richness. To address this problem, we propose to enrich the agent's ontology and extend the traditional conceptualisation of trajectories to provide a more nuanced view of task execution. Structurally Enriched Trajectories (SETs) extend the encoding of sequences of states and their transitions by incorporating hierarchical relations between objects, interactions and affordances. SETs are built as multi-level graphs, providing a detailed representation of the agent dynamics and a transferable functional abstraction of the task. SETs are integrated into an architecture, Structurally Enriched Trajectory Learning and Encoding (SETLE), that employs a heterogeneous graph-based memory structure of multi-level relational dependencies essential for generalisation. Using reinforcement learning as a data generation tool, we demonstrate that SETLE can support downstream tasks, enabling agents to recognise task-relevant structural patterns across diverse environments.
Authors: Yue Su, Xinyu Zhan, Hongjie Fang, Han Xue, Hao-Shu Fang, Yong-Lu Li, Cewu Lu, Lixin Yang
Abstract: Mainstream visuomotor policies predominantly rely on generative models for holistic action prediction, while current autoregressive policies, predicting the next token or chunk, have shown suboptimal results. This motivates a search for more effective learning methods to unleash the potential of autoregressive policies for robotic manipulation. This paper introduces a bidirectionally expanded learning approach, termed Dense Policy, to establish a new paradigm for autoregressive policies in action prediction. It employs a lightweight encoder-only architecture to iteratively unfold the action sequence from an initial single frame into the target sequence in a coarse-to-fine manner with logarithmic-time inference. Extensive experiments validate that our dense policy has superior autoregressive learning capabilities and can surpass existing holistic generative policies. Our policy, example data, and training code will be publicly available upon publication. Project page: https: //selen-suyue.github.io/DspNet/.
Authors: Tong Zhou, Shijin Duan, Gaowen Liu, Charles Fleming, Ramana Rao Kompella, Shaolei Ren, Xiaolin Xu
Abstract: Pre-trained models are valuable intellectual property, capturing both domain-specific and domain-invariant features within their weight spaces. However, model extraction attacks threaten these assets by enabling unauthorized source-domain inference and facilitating cross-domain transfer via the exploitation of domain-invariant features. In this work, we introduce **ProDiF**, a novel framework that leverages targeted weight space manipulation to secure pre-trained models against extraction attacks. **ProDiF** quantifies the transferability of filters and perturbs the weights of critical filters in unsecured memory, while preserving actual critical weights in a Trusted Execution Environment (TEE) for authorized users. A bi-level optimization further ensures resilience against adaptive fine-tuning attacks. Experimental results show that **ProDiF** reduces source-domain accuracy to near-random levels and decreases cross-domain transferability by 74.65\%, providing robust protection for pre-trained models. This work offers comprehensive protection for pre-trained DNN models and highlights the potential of weight space manipulation as a novel approach to model security.
Authors: Akshay Thakur, Matthew J. Zahr
Abstract: The Riemann problem is fundamental in the computational modeling of hyperbolic partial differential equations, enabling the development of stable and accurate upwind schemes. While exact solvers provide robust upwinding fluxes, their high computational cost necessitates approximate solvers. Although approximate solvers achieve accuracy in many scenarios, they produce inaccurate solutions in certain cases. To overcome this limitation, we propose constructing neural network-based surrogate models, trained using supervised learning, designed to map interior and exterior conservative state variables to the corresponding exact flux. Specifically, we propose two distinct approaches: one utilizing a vanilla neural network and the other employing a bi-fidelity neural network. The performance of the proposed approaches is demonstrated through applications to one-dimensional and two-dimensional partial differential equations, showcasing their robustness and accuracy.
Authors: Xiaodi Li, Shaika Chowdhury, Chung Il Wi, Maria Vassilaki, Ken Liu, Terence T Sio, Owen Garrick, Young J Juhn, James R Cerhan, Cui Tao, Nansu Zong
Abstract: Patient matching is the process of linking patients to appropriate clinical trials by accurately identifying and matching their medical records with trial eligibility criteria. We propose LLM-Match, a novel framework for patient matching leveraging fine-tuned open-source large language models. Our approach consists of four key components. First, a retrieval-augmented generation (RAG) module extracts relevant patient context from a vast pool of electronic health records (EHRs). Second, a prompt generation module constructs input prompts by integrating trial eligibility criteria (both inclusion and exclusion criteria), patient context, and system instructions. Third, a fine-tuning module with a classification head optimizes the model parameters using structured prompts and ground-truth labels. Fourth, an evaluation module assesses the fine-tuned model's performance on the testing datasets. We evaluated LLM-Match on four open datasets, n2c2, SIGIR, TREC 2021, and TREC 2022, using open-source models, comparing it against TrialGPT, Zero-Shot, and GPT-4-based closed models. LLM-Match outperformed all baselines.
Authors: Marcello Iotti, Paolo Davini, Jost von Hardenberg, Giuseppe Zappa
Abstract: To this day, accurately simulating local-scale precipitation and reliably reproducing its distribution remains a challenging task. The limited horizontal resolution of Global Climate Models is among the primary factors undermining their skill in this context. The physical mechanisms driving the onset and development of precipitation, especially in extreme events, operate at spatio-temporal scales smaller than those numerically resolved, thus struggling to be captured accurately. In order to circumvent this limitation, several downscaling approaches have been developed over the last decades to address the discrepancy between the spatial resolution of models output and the resolution required by local-scale applications. In this paper, we introduce RainScaleGAN, a conditional deep convolutional Generative Adversarial Network (GAN) for precipitation downscaling. GANs have been effectively used in image super-resolution, an approach highly relevant for downscaling tasks. RainScaleGAN's capabilities are tested in a perfect-model setup, where the spatial resolution of a precipitation dataset is artificially degraded from 0.25$^{\circ}\times$0.25$^{\circ}$ to 2$^{\circ}\times$2$^\circ$, and RainScaleGAN is used to restore it. The developed model outperforms one of the leading precipitation downscaling method found in the literature. RainScaleGAN not only generates a synthetic dataset featuring plausible high-resolution spatial patterns and intensities, but also produces a precipitation distribution with statistics closely mirroring those of the ground-truth dataset. Given that RainScaleGAN's approach is agnostic with respect to the underlying physics, the method has the potential to be applied to other physical variables such as surface winds or temperature.
Authors: Enrico Foglia, Benjamin Bobbia, Nikita Durasov, Michael Bauerheim, Pascal Fua, Stephane Moreau, Thierry Jardin
Abstract: Quantifying model uncertainty is critical for understanding prediction reliability, yet distinguishing between aleatoric and epistemic uncertainty remains challenging. We extend recent work from classification to regression to provide a novel frequentist approach to epistemic and aleatoric uncertainty estimation. We train models to generate conditional predictions by feeding their initial output back as an additional input. This method allows for a rigorous measurement of model uncertainty by observing how prediction responses change when conditioned on the model's previous answer. We provide a complete theoretical framework to analyze epistemic uncertainty in regression in a frequentist way, and explain how it can be exploited in practice to gauge a model's uncertainty, with minimal changes to the original architecture.
Authors: Sang Truong, Yuheng Tu, Percy Liang, Bo Li, Sanmi Koyejo
Abstract: Comprehensive evaluations of language models (LM) during both development and deployment phases are necessary because these models possess numerous capabilities (e.g., mathematical reasoning, legal support, or medical diagnostic) as well as safety risks (e.g., racial bias, toxicity, or misinformation). The average score across a wide range of benchmarks provides a signal that helps guide the use of these LMs in practice. Currently, holistic evaluations are costly due to the large volume of benchmark questions, making frequent evaluations impractical. A popular attempt to lower the cost is to compute the average score on a subset of the benchmark. This approach, unfortunately, often renders an unreliable measure of LM performance because the average score is often confounded with the difficulty of the questions in the benchmark subset. Item response theory (IRT) was designed to address this challenge, providing a reliable measurement by careful controlling for question difficulty. Unfortunately, question difficulty is expensive to estimate. Facing this challenge, we train a model that predicts question difficulty from its content, enabling a reliable measurement at a fraction of the cost. In addition, we leverage this difficulty predictor to further improve the evaluation efficiency through training a question generator given a difficulty level. This question generator is essential in adaptive testing, where, instead of using a random subset of the benchmark questions, informative questions are adaptively chosen based on the current estimation of LLM performance. Experiments on 22 common natural language benchmarks and 172 LMs show that this approach is more reliable and efficient compared to current common practice.
Authors: Ewan R. S. Wallace, Nathan C. Frey, Joshua A. Rackers
Abstract: Ligand strain energy, the energy difference between the bound and unbound conformations of a ligand, is an important component of structure-based small molecule drug design. A large majority of observed ligands in protein-small molecule co-crystal structures bind in low-strain conformations, making strain energy a useful filter for structure-based drug design. In this work we present a tool for calculating ligand strain with a high accuracy. StrainRelief uses a MACE Neural Network Potential (NNP), trained on a large database of Density Functional Theory (DFT) calculations to estimate ligand strain of neutral molecules with quantum accuracy. We show that this tool estimates strain energy differences relative to DFT to within 1.4 kcal/mol, more accurately than alternative NNPs. These results highlight the utility of NNPs in drug discovery, and provide a useful tool for drug discovery teams.
Authors: Hai-Long Sun, Zhun Sun, Houwen Peng, Han-Jia Ye
Abstract: Recent advancements in Large Language Models (LLMs) have demonstrated enhanced reasoning capabilities, evolving from Chain-of-Thought (CoT) prompting to advanced, product-oriented solutions like OpenAI o1. During our re-implementation of this model, we noticed that in multimodal tasks requiring visual input (e.g., geometry problems), Multimodal LLMs (MLLMs) struggle to maintain focus on the visual information, in other words, MLLMs suffer from a gradual decline in attention to visual information as reasoning progresses, causing text-over-relied outputs. To investigate this, we ablate image inputs during long-chain reasoning. Concretely, we truncate the reasoning process midway, then re-complete the reasoning process with the input image removed. We observe only a ~2% accuracy drop on MathVista's test-hard subset, revealing the model's textual outputs dominate the following reasoning process. Motivated by this, we propose Take-along Visual Conditioning (TVC), a strategy that shifts image input to critical reasoning stages and compresses redundant visual tokens via dynamic pruning. This methodology helps the model retain attention to the visual components throughout the reasoning. Our approach achieves state-of-the-art performance on average across five mathematical reasoning benchmarks (+3.4% vs previous sota), demonstrating the effectiveness of TVC in enhancing multimodal reasoning systems.
Authors: Mengyao Lyu, Yan Li, Huasong Zhong, Wenhao Yang, Hui Chen, Jungong Han, Guiguang Ding, Zhenheng Yang
Abstract: The hypothesis that pretrained large language models (LLMs) necessitate only minimal supervision during the fine-tuning (SFT) stage (Zhou et al., 2024) has been substantiated by recent advancements in data curation and selection research. However, their stability and generalizability are compromised due to the vulnerability to experimental setups and validation protocols, falling short of surpassing random sampling (Diddee & Ippolito, 2024; Xia et al., 2024b). Built upon LLMs, multi-modal LLMs (MLLMs), combined with the sheer token volume and heightened heterogeneity of data sources, amplify both the significance and complexity of data selection. To harvest multi-modal instructional data in a robust and efficient manner, we re-define the granularity of the quality metric by decomposing it into 14 vision-language-related capabilities, and introduce multi-modal rich scorers to evaluate the capabilities of each data candidate. To promote diversity, in light of the inherent objective of the alignment stage, we take interaction style as diversity indicator and use a multi-modal rich styler to identify data instruction patterns. In doing so, our multi-modal rich scorers and styler (mmSSR) guarantee that high-scoring information is conveyed to users in diversified forms. Free from embedding-based clustering or greedy sampling, mmSSR efficiently scales to millions of data with varying budget constraints, supports customization for general or specific capability acquisition, and facilitates training-free generalization to new domains for curation. Across 10+ experimental settings, validated by 14 multi-modal benchmarks, we demonstrate consistent improvements over random sampling, baseline strategies and state-of-the-art selection methods, achieving 99.1% of full performance with only 30% of the 2.6M data.
Authors: Qing Zhou, Junyu Gao, Qi Wang
Abstract: The rapid growth of dataset scales has been a key driver in advancing deep learning research. However, as dataset scale increases, the training process becomes increasingly inefficient due to the presence of low-value samples, including excessive redundant samples, overly challenging samples, and inefficient easy samples that contribute little to model improvement.To address this challenge, we propose Scale Efficient Training (SeTa) for large datasets, a dynamic sample pruning approach that losslessly reduces training time. To remove low-value samples, SeTa first performs random pruning to eliminate redundant samples, then clusters the remaining samples according to their learning difficulty measured by loss. Building upon this clustering, a sliding window strategy is employed to progressively remove both overly challenging and inefficient easy clusters following an easy-to-hard curriculum.We conduct extensive experiments on large-scale synthetic datasets, including ToCa, SS1M, and ST+MJ, each containing over 3 million samples.SeTa reduces training costs by up to 50\% while maintaining or improving performance, with minimal degradation even at 70\% cost reduction. Furthermore, experiments on various scale real datasets across various backbones (CNNs, Transformers, and Mambas) and diverse tasks (instruction tuning, multi-view stereo, geo-localization, composed image retrieval, referring image segmentation) demonstrate the powerful effectiveness and universality of our approach. Code is available at https://github.com/mrazhou/SeTa.
Authors: James Burgess, Jeffrey J Nirschl, Laura Bravo-S\'anchez, Alejandro Lozano, Sanket Rajan Gupte, Jesus G. Galaz-Montoya, Yuhui Zhang, Yuchang Su, Disha Bhowmik, Zachary Coman, Sarina M. Hasan, Alexandra Johannesson, William D. Leineweber, Malvika G Nair, Ridhi Yarlagadda, Connor Zuraski, Wah Chiu, Sarah Cohen, Jan N. Hansen, Manuel D Leonetti, Chad Liu, Emma Lundberg, Serena Yeung-Levy
Abstract: Scientific research demands sophisticated reasoning over multimodal data, a challenge especially prevalent in biology. Despite recent advances in multimodal large language models (MLLMs) for AI-assisted research, existing multimodal reasoning benchmarks only target up to college-level difficulty, while research-level benchmarks emphasize lower-level perception, falling short of the complex multimodal reasoning needed for scientific discovery. To bridge this gap, we introduce MicroVQA, a visual-question answering (VQA) benchmark designed to assess three reasoning capabilities vital in research workflows: expert image understanding, hypothesis generation, and experiment proposal. MicroVQA consists of 1,042 multiple-choice questions (MCQs) curated by biology experts across diverse microscopy modalities, ensuring VQA samples represent real scientific practice. In constructing the benchmark, we find that standard MCQ generation methods induce language shortcuts, motivating a new two-stage pipeline: an optimized LLM prompt structures question-answer pairs into MCQs; then, an agent-based `RefineBot' updates them to remove shortcuts. Benchmarking on state-of-the-art MLLMs reveal a peak performance of 53\%; models with smaller LLMs only slightly underperform top models, suggesting that language-based reasoning is less challenging than multimodal reasoning; and tuning with scientific articles enhances performance. Expert analysis of chain-of-thought responses shows that perception errors are the most frequent, followed by knowledge errors and then overgeneralization errors. These insights highlight the challenges in multimodal scientific reasoning, showing MicroVQA is a valuable resource advancing AI-driven biomedical research. MicroVQA is available at https://huggingface.co/datasets/jmhb/microvqa, and project page at https://jmhb0.github.io/microvqa.
URLs: https://huggingface.co/datasets/jmhb/microvqa,, https://jmhb0.github.io/microvqa.
Authors: Cheoljoon Jeong, Xubo Yue, Seokhyun Chung
Abstract: Many failure mechanisms of machinery are closely related to the behavior of condition monitoring (CM) signals. To achieve a cost-effective preventive maintenance strategy, accurate remaining useful life (RUL) prediction based on the signals is of paramount importance. However, the CM signals are often recorded at different factories and production lines, with limited amounts of data. Unfortunately, these datasets have rarely been shared between the sites due to data confidentiality and ownership issues, a lack of computing and storage power, and high communication costs associated with data transfer between sites and a data center. Another challenge in real applications is that the CM signals are often not explicitly specified \textit{a priori}, meaning that existing methods, which often usually a parametric form, may not be applicable. To address these challenges, we propose a new prognostic framework for RUL prediction using the joint modeling of nonlinear degradation signals and time-to-failure data within a federated learning scheme. The proposed method constructs a nonparametric degradation model using a federated multi-output Gaussian process and then employs a federated survival model to predict failure times and probabilities for in-service machinery. The superiority of the proposed method over other alternatives is demonstrated through comprehensive simulation studies and a case study using turbofan engine degradation signal data that include run-to-failure events.
Authors: Alisa Liu, Jonathan Hayase, Valentin Hofmann, Sewoong Oh, Noah A. Smith, Yejin Choi
Abstract: The assumption across nearly all language model (LM) tokenization schemes is that tokens should be subwords, i.e., contained within word boundaries. While providing a seemingly reasonable inductive bias, is this common practice limiting the potential of modern LMs? Whitespace is not a reliable delimiter of meaning, as evidenced by multi-word expressions (e.g., "by the way"), crosslingual variation in the number of words needed to express a concept (e.g., "spacesuit helmet" in German is "raumanzughelm"), and languages that do not use whitespace at all (e.g., Chinese). To explore the potential of tokenization beyond subwords, we introduce a "superword" tokenizer, SuperBPE, which incorporates a simple pretokenization curriculum into the byte-pair encoding (BPE) algorithm to first learn subwords, then superwords that bridge whitespace. This brings dramatic improvements in encoding efficiency: when fixing the vocabulary size to 200k, SuperBPE encodes a fixed piece of text with up to 33% fewer tokens than BPE on average. In experiments, we pretrain 8B transformer LMs from scratch while fixing the model size, vocabulary size, and train compute, varying *only* the algorithm for learning the vocabulary. Our model trained with SuperBPE achieves an average +4.0% absolute improvement over the BPE baseline across 30 downstream tasks (including +8.2% on MMLU), while simultaneously requiring 27% less compute at inference time. In analysis, we find that SuperBPE results in segmentations of text that are more uniform in per-token difficulty. Qualitatively, this may be because SuperBPE tokens often capture common multi-word expressions that function semantically as a single unit. SuperBPE is a straightforward, local modification to tokenization that improves both encoding efficiency and downstream performance, yielding better language models overall.
Authors: Thomas Monninger, Md Zafar Anwar, Stanislaw Antol, Steffen Staab, Sihao Ding
Abstract: Autonomous driving requires an understanding of the infrastructure elements, such as lanes and crosswalks. To navigate safely, this understanding must be derived from sensor data in real-time and needs to be represented in vectorized form. Learned Bird's-Eye View (BEV) encoders are commonly used to combine a set of camera images from multiple views into one joint latent BEV grid. Traditionally, from this latent space, an intermediate raster map is predicted, providing dense spatial supervision but requiring post-processing into the desired vectorized form. More recent models directly derive infrastructure elements as polylines using vectorized map decoders, providing instance-level information. Our approach, Augmentation Map Network (AugMapNet), proposes latent BEV grid augmentation, a novel technique that significantly enhances the latent BEV representation. AugMapNet combines vector decoding and dense spatial supervision more effectively than existing architectures while remaining as straightforward to integrate and as generic as auxiliary supervision. Experiments on nuScenes and Argoverse2 datasets demonstrate significant improvements in vectorized map prediction performance up to 13.3% over the StreamMapNet baseline on 60m range and greater improvements on larger ranges. We confirm transferability by applying our method to another baseline and find similar improvements. A detailed analysis of the latent BEV grid confirms a more structured latent space of AugMapNet and shows the value of our novel concept beyond pure performance improvement. The code will be released soon.
Authors: Lijie Fan, Luming Tang, Siyang Qin, Tianhong Li, Xuan Yang, Siyuan Qiao, Andreas Steiner, Chen Sun, Yuanzhen Li, Tao Zhu, Michael Rubinstein, Michalis Raptis, Deqing Sun, Radu Soricut
Abstract: We present UniFluid, a unified autoregressive framework for joint visual generation and understanding leveraging continuous visual tokens. Our unified autoregressive architecture processes multimodal image and text inputs, generating discrete tokens for text and continuous tokens for image. We find though there is an inherent trade-off between the image generation and understanding task, a carefully tuned training recipe enables them to improve each other. By selecting an appropriate loss balance weight, the unified model achieves results comparable to or exceeding those of single-task baselines on both tasks. Furthermore, we demonstrate that employing stronger pre-trained LLMs and random-order generation during training is important to achieve high-fidelity image generation within this unified framework. Built upon the Gemma model series, UniFluid exhibits competitive performance across both image generation and understanding, demonstrating strong transferability to various downstream tasks, including image editing for generation, as well as visual captioning and question answering for understanding.
Authors: Qin Liu, Wenxuan Zhou, Nan Xu, James Y. Huang, Fei Wang, Sheng Zhang, Hoifung Poon, Muhao Chen
Abstract: One critical challenge for large language models (LLMs) for making complex reasoning is their reliance on matching reasoning patterns from training data, instead of proactively selecting the most appropriate cognitive strategy to solve a given task. Existing approaches impose fixed cognitive structures that enhance performance in specific tasks but lack adaptability across diverse scenarios. To address this limitation, we introduce METASCALE, a test-time scaling framework based on meta-thoughts -- adaptive thinking strategies tailored to each task. METASCALE initializes a pool of candidate meta-thoughts, then iteratively selects and evaluates them using a multi-armed bandit algorithm with upper confidence bound selection, guided by a reward model. To further enhance adaptability, a genetic algorithm evolves high-reward meta-thoughts, refining and extending the strategy pool over time. By dynamically proposing and optimizing meta-thoughts at inference time, METASCALE improves both accuracy and generalization across a wide range of tasks. Experimental results demonstrate that MetaScale consistently outperforms standard inference approaches, achieving an 11% performance gain in win rate on Arena-Hard for GPT-4o, surpassing o1-mini by 0.9% under style control. Notably, METASCALE scales more effectively with increasing sampling budgets and produces more structured, expert-level responses.
Authors: Dustin L. Arendt, Matthew Broussard, Bala Krishnamoorthy, Nathaniel Saul, Amber Thrall
Abstract: We define a new filtration called the Steinhaus filtration built from a single cover based on a generalized Steinhaus distance, a generalization of Jaccard distance. The homology persistence module of a Steinhaus filtration with infinitely many cover elements may not be $q$-tame, even when the covers are in a totally bounded space. While this may pose a challenge to derive stability results, we show that the Steinhaus filtration is stable when the cover is finite. We show that while the \v{C}ech and Steinhaus filtrations are not isomorphic in general, they are isomorphic for a finite point set in dimension one. Furthermore, the VR filtration completely determines the $1$-skeleton of the Steinhaus filtration in arbitrary dimension. We then develop a language and theory for stable paths within the Steinhaus filtration. We demonstrate how the framework can be applied to several applications where a standard metric may not be defined but a cover is readily available. We introduce a new perspective for modeling recommendation system datasets. As an example, we look at a movies dataset and we find the stable paths identified in our framework represent a sequence of movies constituting a gentle transition and ordering from one genre to another. For explainable machine learning, we apply the Mapper algorithm for model induction by building a filtration from a single Mapper complex, and provide explanations in the form of stable paths between subpopulations. For illustration, we build a Mapper complex from a supervised machine learning model trained on the FashionMNIST dataset. Stable paths in the Steinhaus filtration provide improved explanations of relationships between subpopulations of images.
Authors: Satyapriya Krishna, Tessa Han, Alex Gu, Steven Wu, Shahin Jabbari, Himabindu Lakkaraju
Abstract: As various post hoc explanation methods are increasingly being leveraged to explain complex models in high-stakes settings, it becomes critical to develop a deeper understanding of if and when the explanations output by these methods disagree with each other, and how such disagreements are resolved in practice. However, there is little to no research that provides answers to these critical questions. In this work, we introduce and study the disagreement problem in explainable machine learning. More specifically, we formalize the notion of disagreement between explanations, analyze how often such disagreements occur in practice, and how practitioners resolve these disagreements. We first conduct interviews with data scientists to understand what constitutes disagreement between explanations generated by different methods for the same model prediction and introduce a novel quantitative framework to formalize this understanding. We then leverage this framework to carry out a rigorous empirical analysis with four real-world datasets, six state-of-the-art post hoc explanation methods, and six different predictive models, to measure the extent of disagreement between the explanations generated by various popular explanation methods. In addition, we carry out an online user study with data scientists to understand how they resolve the aforementioned disagreements. Our results indicate that (1) state-of-the-art explanation methods often disagree in terms of the explanations they output, and (2) machine learning practitioners often employ ad hoc heuristics when resolving such disagreements. These findings suggest that practitioners may be relying on misleading explanations when making consequential decisions. They also underscore the importance of developing principled frameworks for effectively evaluating and comparing explanations output by various explanation techniques.
Authors: Kevin Han Huang, Peter Orbanz, Morgane Austern
Abstract: We provide universality results that quantify how data augmentation affects the variance and limiting distribution of estimates through simple surrogates, and analyze several specific models in detail. The results confirm some observations made in machine learning practice, but also lead to unexpected findings: Data augmentation may increase rather than decrease the uncertainty of estimates, such as the empirical prediction risk. It can act as a regularizer, but fails to do so in certain high-dimensional problems, and it may shift the double-descent peak of an empirical risk. Overall, the analysis shows that several properties data augmentation has been attributed with are not either true or false, but rather depend on a combination of factors -- notably the data distribution, the properties of the estimator, and the interplay of sample size, number of augmentations, and dimension. As our main theoretical tool, we develop an adaptation of Lindeberg's technique for block dependence. The resulting universality regime may be Gaussian or non-Gaussian.
Authors: Zixuan Dong, Che Wang, Keith Ross
Abstract: In reinforcement learning, Monte Carlo algorithms update the Q function by averaging the episodic returns. In the Monte Carlo UCB (MC-UCB) algorithm, the action taken in each state is the action that maximizes the Q function plus an Upper Confidence Bounds (UCB) exploration term, which biases the choice of actions to those that have been chosen less frequently. Although there has been significant work on establishing regret bounds for MC-UCB, most of that work has been focused on finite-horizon versions of the problem, for which each episode terminates after a constant number of steps. For such finite-horizon problems, the optimal policy depends both on the current state and the time within the episode. However, for many natural episodic problems, such as games like Go and Chess and robotic tasks, the episode is of random length and the optimal policy is stationary. For such environments, it is an open question whether the Q-function in MC-UCB will converge to the optimal Q function; we conjecture that, unlike Q-learning, it does not converge for all MDPs. We nevertheless show that for a large class of MDPs, which includes stochastic MDPs such as blackjack and deterministic MDPs such as Go, the Q function in MC-UCB converges almost surely to the optimal Q function. An immediate corollary of this result is that it also converges almost surely for all finite-horizon MDPs. We also provide numerical experiments, providing further insights into MC-UCB.
Authors: Han Wu, Sareh Rowlands, Johan Wahlstrom
Abstract: As cloud computing becomes pervasive, deep learning models are deployed on cloud servers and then provided as APIs to end users. However, black-box adversarial attacks can fool image classification models without access to model structure and weights. Recent studies have reported attack success rates of over 95% with fewer than 1,000 queries. Then the question arises: whether black-box attacks have become a real threat against cloud APIs? To shed some light on this, our research indicates that black-box attacks are not as effective against cloud APIs as proposed in research papers due to several common mistakes that overestimate the efficiency of black-box attacks. To avoid similar mistakes, we conduct black-box attacks directly on cloud APIs rather than local models.
Authors: Javier Marin
Abstract: This work proposes a method to evaluate synthetic tabular data generated to augment small sample datasets. While data augmentation techniques can increase sample counts for machine learning applications, traditional validation approaches fail when applied to extremely limited sample sizes. Our experiments across four datasets reveal significant inconsistencies between global metrics and topological measures, with statistical tests producing unreliable significance values due to insufficient sample sizes. We demonstrate that common metrics like propensity scoring and MMD often suggest similarity where fundamental topological differences exist. Our proposed normalized Bottleneck distance based metric provides complementary insights but suffers from high variability across experimental runs and occasional values exceeding theoretical bounds, showing inherent instability in topological approaches for very small datasets. These findings highlight the critical need for multi-faceted evaluation methodologies when validating synthetic data generated from limited samples, as no single metric reliably captures both distributional and structural similarity.
Authors: Shujian Yu, Hongming Li, Sigurd L{\o}kse, Robert Jenssen, Jos\'e C. Pr\'incipe
Abstract: The Cauchy-Schwarz (CS) divergence was developed by Pr\'{i}ncipe et al. in 2000. In this paper, we extend the classic CS divergence to quantify the closeness between two conditional distributions and show that the developed conditional CS divergence can be simply estimated by a kernel density estimator from given samples. We illustrate the advantages (e.g., rigorous faithfulness guarantee, lower computational complexity, higher statistical power, and much more flexibility in a wide range of applications) of our conditional CS divergence over previous proposals, such as the conditional KL divergence and the conditional maximum mean discrepancy. We also demonstrate the compelling performance of conditional CS divergence in two machine learning tasks related to time series data and sequential inference, namely time series clustering and uncertainty-guided exploration for sequential decision making. The code of conditional CS divergence is available at https://github.com/SJYuCNEL/conditional_CS_divergence.
URLs: https://github.com/SJYuCNEL/conditional_CS_divergence.
Authors: Zhihao Shi, Xize Liang, Jie Wang
Abstract: The message passing-based graph neural networks (GNNs) have achieved great success in many real-world applications. However, training GNNs on large-scale graphs suffers from the well-known neighbor explosion problem, i.e., the exponentially increasing dependencies of nodes with the number of message passing layers. Subgraph-wise sampling methods -- a promising class of mini-batch training techniques -- discard messages outside the mini-batches in backward passes to avoid the neighbor explosion problem at the expense of gradient estimation accuracy. This poses significant challenges to their convergence analysis and convergence speeds, which seriously limits their reliable real-world applications. To address this challenge, we propose a novel subgraph-wise sampling method with a convergence guarantee, namely Local Message Compensation (LMC). To the best of our knowledge, LMC is the {\it first} subgraph-wise sampling method with provable convergence. The key idea of LMC is to retrieve the discarded messages in backward passes based on a message passing formulation of backward passes. By efficient and effective compensations for the discarded messages in both forward and backward passes, LMC computes accurate mini-batch gradients and thus accelerates convergence. We further show that LMC converges to first-order stationary points of GNNs. Experiments on large-scale benchmark tasks demonstrate that LMC significantly outperforms state-of-the-art subgraph-wise sampling methods in terms of efficiency.
Authors: Kun Yi, Qi Zhang, Wei Fan, Longbing Cao, Shoujin Wang, Guodong Long, Liang Hu, Hui He, Qingsong Wen, Hui Xiong
Abstract: Recently, frequency transformation (FT) has been increasingly incorporated into deep learning models to significantly enhance state-of-the-art accuracy and efficiency in time series analysis. The advantages of FT, such as high efficiency and a global view, have been rapidly explored and exploited in various time series tasks and applications, demonstrating the promising potential of FT as a new deep learning paradigm for time series analysis. Despite the growing attention and the proliferation of research in this emerging field, there is currently a lack of a systematic review and in-depth analysis of deep learning-based time series models with FT. It is also unclear why FT can enhance time series analysis and what its limitations are in the field. To address these gaps, we present a comprehensive review that systematically investigates and summarizes the recent research advancements in deep learning-based time series analysis with FT. Specifically, we explore the primary approaches used in current models that incorporate FT, the types of neural networks that leverage FT, and the representative FT-equipped models in deep time series analysis. We propose a novel taxonomy to categorize the existing methods in this field, providing a structured overview of the diverse approaches employed in incorporating FT into deep learning models for time series analysis. Finally, we highlight the advantages and limitations of FT for time series modeling and identify potential future research directions that can further contribute to the community of time series analysis.
Authors: Maolin Wang, Yu Pan, Zenglin Xu, Guangxi Li, Xiangli Yang, Danilo Mandic, Andrzej Cichocki
Abstract: Tensor networks (TNs) and neural networks (NNs) are two fundamental data modeling approaches. TNs were introduced to solve the curse of dimensionality in large-scale tensors by converting an exponential number of dimensions to polynomial complexity. As a result, they have attracted significant attention in the fields of quantum physics and machine learning. Meanwhile, NNs have displayed exceptional performance in various applications, e.g., computer vision, natural language processing, and robotics research. Interestingly, although these two types of networks originate from different observations, they are inherently linked through the typical multilinearity structure underlying both TNs and NNs, thereby motivating a significant number of developments regarding combinations of TNs and NNs. In this paper, we refer to these combinations as tensorial neural networks~(TNNs) and present an introduction to TNNs from both data processing and model architecture perspectives. From the data perspective, we explore the capabilities of TNNs in multi-source fusion, multimodal pooling, data compression, multi-task training, and quantum data processing. From the model perspective, we examine TNNs' integration with various architectures, including Convolutional Neural Networks, Recurrent Neural Networks, Graph Neural Networks, Transformers, Large Language Models, and Quantum Neural Networks. Furthermore, this survey also explores methods for improving TNNs, examines flexible toolboxes for implementing TNNs, and documents TNN development while highlighting potential future directions. To the best of our knowledge, this is the first comprehensive survey that bridges the connections among NNs and TNs. We provide a curated list of TNNs at https://github.com/tnbar/awesome-tensorial-neural-networks.
URLs: https://github.com/tnbar/awesome-tensorial-neural-networks.
Authors: Farhad Mortezapour Shiri, Thinagaran Perumal, Norwati Mustapha, Raihani Mohamed
Abstract: Deep learning (DL) has emerged as a powerful subset of machine learning (ML) and artificial intelligence (AI), outperforming traditional ML methods, especially in handling unstructured and large datasets. Its impact spans across various domains, including speech recognition, healthcare, autonomous vehicles, cybersecurity, predictive analytics, and more. However, the complexity and dynamic nature of real-world problems present challenges in designing effective deep learning models. Consequently, several deep learning models have been developed to address different problems and applications. In this article, we conduct a comprehensive survey of various deep learning models, including Convolutional Neural Network (CNN), Recurrent Neural Network (RNN), Temporal Convolutional Networks (TCN), Transformer, Kolmogorov-Arnold networks (KAN), Generative Models, Deep Reinforcement Learning (DRL), and Deep Transfer Learning. We examine the structure, applications, benefits, and limitations of each model. Furthermore, we perform an analysis using three publicly available datasets: IMDB, ARAS, and Fruit-360. We compared the performance of six renowned deep learning models: CNN, RNN, Long Short-Term Memory (LSTM), Bidirectional LSTM, Gated Recurrent Unit (GRU), and Bidirectional GRU alongside two newer models, TCN and Transformer, using the IMDB and ARAS datasets. Additionally, we evaluated the performance of eight CNN-based models, including VGG (Visual Geometry Group), Inception, ResNet (Residual Network), InceptionResNet, Xception (Extreme Inception), MobileNet, DenseNet (Dense Convolutional Network), and NASNet (Neural Architecture Search Network), for image classification tasks using the Fruit-360 dataset.
Authors: Nikolaos Louloudakis, Perry Gibson, Jos\'e Cano, Ajitha Rajan
Abstract: As the usage of Artificial Intelligence (AI) on resource-intensive and safety-critical tasks increases, a variety of Machine Learning (ML) compilers have been developed, enabling compatibility of Deep Neural Networks (DNNs) with a variety of hardware acceleration devices. However, given that DNNs are widely utilized for challenging and demanding tasks, the behavior of these compilers must be verified. To this direction, we propose MutateNN, a tool that utilizes elements of both differential and mutation testing in order to examine the robustness of image recognition models when deployed on hardware accelerators with different capabilities, in the presence of faults in their target device code - introduced either by developers, or problems in their compilation process. We focus on the image recognition domain by applying mutation testing to 7 well-established DNN models, introducing 21 mutations of 6 different categories. We deployed our mutants on 4 different hardware acceleration devices of varying capabilities and observed that DNN models presented discrepancies of up to 90.3% in mutants related to conditional operators across devices. We also observed that mutations related to layer modification, arithmetic types and input affected severely the overall model performance (up to 99.8%) or led to model crashes, in a consistent manner across devices.
Authors: Shan Sha, Shenglong Zhou, Lingchen Kong, Geoffrey Ye Li
Abstract: Decentralized Federated Learning (DFL) enables collaborative model training without a central server but faces challenges in efficiency, stability, and trustworthiness due to communication and computational limitations among distributed nodes. To address these critical issues, we introduce a sparsity constraint on the shared model, leading to Sparse DFL (SDFL), and propose a novel algorithm, CEPS. The sparsity constraint facilitates the use of one-bit compressive sensing to transmit one-bit information between partially selected neighbour nodes at specific steps, thereby significantly improving communication efficiency. Moreover, we integrate differential privacy into the algorithm to ensure privacy preservation and bolster the trustworthiness of the learning process. Furthermore, CEPS is underpinned by theoretical guarantees regarding both convergence and privacy. Numerical experiments validate the effectiveness of the proposed algorithm in improving communication and computation efficiency while maintaining a high level of trustworthiness.
Authors: Yannick Eich, Bastian Alt, Heinz Koeppl
Abstract: We propose a novel, tractable latent state inference scheme for Markov jump processes, for which exact inference is often intractable. Our approach is based on an entropic matching framework that can be embedded into the well-known expectation propagation algorithm. We demonstrate the effectiveness of our method by providing closed-form results for a simple family of approximate distributions and apply it to the general class of chemical reaction networks, which are a crucial tool for modeling in systems biology. Moreover, we derive closed-form expressions for point estimation of the underlying parameters using an approximate expectation maximization procedure. We evaluate our method across various chemical reaction networks and compare it to multiple baseline approaches, demonstrating superior performance in approximating the mean of the posterior process. Finally, we discuss the limitations of our method and potential avenues for future improvement, highlighting its promising direction for addressing complex continuous-time Bayesian inference problems.
Authors: Yida Yin, Zhiqiu Xu, Zhiyuan Li, Trevor Darrell, Zhuang Liu
Abstract: Stochastic Variance Reduced Gradient (SVRG), introduced by Johnson & Zhang (2013), is a theoretically compelling optimization method. However, as Defazio & Bottou (2019) highlight, its effectiveness in deep learning is yet to be proven. In this work, we demonstrate the potential of SVRG in optimizing real-world neural networks. Our empirical analysis finds that, for deeper neural networks, the strength of the variance reduction term in SVRG should be smaller and decrease as training progresses. Inspired by this, we introduce a multiplicative coefficient $\alpha$ to control the strength and adjust it through a linear decay schedule. We name our method $\alpha$-SVRG. Our results show $\alpha$-SVRG better optimizes models, consistently reducing training loss compared to the baseline and standard SVRG across various model architectures and multiple image classification datasets. We hope our findings encourage further exploration into variance reduction techniques in deep learning. Code is available at github.com/davidyyd/alpha-SVRG.
Authors: Sungbin Shin, Dongyeop Lee, Maksym Andriushchenko, Namhoon Lee
Abstract: Training overparameterized neural networks often yields solutions with varying generalization capabilities, even when achieving similar training losses. Recent evidence indicates a strong correlation between the sharpness of a minimum and its generalization error, leading to increased interest in optimization methods that explicitly seek flatter minima for improved generalization. Despite its contemporary relevance to overparameterization, however, this sharpness-aware minimization (SAM) strategy has not been studied much yet as to exactly how it is affected by overparameterization. In this work, we analyze SAM under varying degrees of overparameterization, presenting both empirical and theoretical findings that reveal its critical influence on SAM's effectiveness. First, we conduct extensive numerical experiments across diverse domains, demonstrating that SAM consistently benefits from overparameterization. Next, we attribute this phenomenon to the interplay between the enlarged solution space and increased implicit bias resulting from overparameterization. Furthermore, we show that this effect is particularly pronounced in practical settings involving label noise and sparsity, and yet, sufficient regularization is necessary. Last but not least, we provide other theoretical insights into how overparameterization helps SAM achieve minima with more uniform Hessian moments compared to SGD, and much faster convergence at a linear rate.
Authors: Sujay Nagaraj, Walter Gerych, Sana Tonekaboni, Anna Goldenberg, Berk Ustun, Thomas Hartvigsen
Abstract: Many time series classification tasks, where labels vary over time, are affected by label noise that also varies over time. Such noise can cause label quality to improve, worsen, or periodically change over time. We first propose and formalize temporal label noise, an unstudied problem for sequential classification of time series. In this setting, multiple labels are recorded over time while being corrupted by a time-dependent noise function. We first demonstrate the importance of modeling the temporal nature of the label noise function and how existing methods will consistently underperform. We then propose methods to train noise-tolerant classifiers by estimating the temporal label noise function directly from data. We show that our methods lead to state-of-the-art performance under diverse types of temporal label noise on real-world datasets
Authors: Christophe Roux, Max Zimmer, Sebastian Pokutta
Abstract: Federated Learning (FL) algorithms using Knowledge Distillation (KD) have received increasing attention due to their favorable properties with respect to privacy, non-i.i.d. data and communication cost. These methods depart from transmitting model parameters and instead communicate information about a learning task by sharing predictions on a public dataset. In this work, we study the performance of such approaches in the byzantine setting, where a subset of the clients act in an adversarial manner aiming to disrupt the learning process. We show that KD-based FL algorithms are remarkably resilient and analyze how byzantine clients can influence the learning process. Based on these insights, we introduce two new byzantine attacks and demonstrate their ability to break existing byzantine-resilient methods. Additionally, we propose a novel defence method which enhances the byzantine resilience of KD-based FL algorithms. Finally, we provide a general framework to obfuscate attacks, making them significantly harder to detect, thereby improving their effectiveness. Our findings serve as an important building block in the analysis of byzantine FL, contributing through the development of new attacks and new defence mechanisms, further advancing the robustness of KD-based FL algorithms.
Authors: Hao Chen, Zihan Wang, Ran Tao, Hongxin Wei, Xing Xie, Masashi Sugiyama, Bhiksha Raj, Jindong Wang
Abstract: Foundation models are usually pre-trained on large-scale datasets and then adapted to downstream tasks through tuning. However, the large-scale pre-training datasets, often inaccessible or too expensive to handle, can contain label noise that may adversely affect the generalization of the model and pose unexpected risks. This paper stands out as the first work to comprehensively understand and analyze the nature of noise in pre-training datasets and then effectively mitigate its impacts on downstream tasks. Specifically, through extensive experiments of fully-supervised and image-text contrastive pre-training on synthetic noisy ImageNet-1K, YFCC15M, and CC12M datasets, we demonstrate that, while slight noise in pre-training can benefit in-domain (ID) performance, where the training and testing data share a similar distribution, it always deteriorates out-of-domain (OOD) performance, where training and testing distributions are significantly different. These observations are agnostic to scales of pre-training datasets, pre-training noise types, model architectures, pre-training objectives, downstream tuning methods, and downstream applications. We empirically ascertain that the reason behind this is that the pre-training noise shapes the feature space differently. We then propose a tuning method (NMTune) to affine the feature space to mitigate the malignant effect of noise and improve generalization, which is applicable in both parameter-efficient and black-box tuning manners. We additionally conduct extensive experiments on popular vision and language models, including APIs, which are supervised and self-supervised pre-trained on realistic noisy data for evaluation. Our analysis and results demonstrate the importance of this novel and fundamental research direction, which we term as Noisy Model Learning.
Authors: Yiping Ji, Hemanth Saratchandran, Cameron Gordon, Zeyu Zhang, Simon Lucey
Abstract: Low-rank decomposition has emerged as a vital tool for enhancing parameter efficiency in neural network architectures, gaining traction across diverse applications in machine learning. These techniques significantly lower the number of parameters, striking a balance between compactness and performance. However, a common challenge has been the compromise between parameter efficiency and the accuracy of the model, where reduced parameters often lead to diminished accuracy compared to their full-rank counterparts. In this work, we propose a novel theoretical framework that integrates a sinusoidal function within the low-rank decomposition process. This approach not only preserves the benefits of the parameter efficiency characteristic of low-rank methods but also increases the decomposition's rank, thereby enhancing model performance. Our method proves to be a plug in enhancement for existing low-rank models, as evidenced by its successful application in Vision Transformers (ViT), Large Language Models (LLMs), Neural Radiance Fields (NeRF) and 3D shape modelling.
Authors: Ramansh Sharma, Varun Shankar
Abstract: We present a novel deep operator network (DeepONet) architecture for operator learning, the ensemble DeepONet, that allows for enriching the trunk network of a single DeepONet with multiple distinct trunk networks. This trunk enrichment allows for greater expressivity and generalization capabilities over a range of operator learning problems. We also present a spatial mixture-of-experts (MoE) DeepONet trunk network architecture that utilizes a partition-of-unity (PoU) approximation to promote spatial locality and model sparsity in the operator learning problem. We first prove that both the ensemble and PoU-MoE DeepONets are universal approximators. We then demonstrate that ensemble DeepONets containing a trunk ensemble of a standard trunk, the PoU-MoE trunk, and/or a proper orthogonal decomposition (POD) trunk can achieve 2-4x lower relative $\ell_2$ errors than standard DeepONets and POD-DeepONets on both standard and challenging new operator learning problems involving partial differential equations (PDEs) in two and three dimensions. Our new PoU-MoE formulation provides a natural way to incorporate spatial locality and model sparsity into any neural network architecture, while our new ensemble DeepONet provides a powerful and general framework for incorporating basis enrichment in scientific machine learning architectures for operator learning.
Authors: Simon Weissmann, Sara Klein, Wa\"iss Azizian, Leif D\"oring
Abstract: Stochastic gradient methods are among the most important algorithms in training machine learning problems. While classical assumptions such as strong convexity allow a simple analysis they are rarely satisfied in applications. In recent years, global and local gradient domination properties have shown to be a more realistic replacement of strong convexity. They were proved to hold in diverse settings such as (simple) policy gradient methods in reinforcement learning and training of deep neural networks with analytic activation functions. We prove almost sure convergence rates $f(X_n)-f^*\in o\big( n^{-\frac{1}{4\beta-1}+\epsilon}\big)$ of the last iterate for stochastic gradient descent (with and without momentum) under global and local $\beta$-gradient domination assumptions. The almost sure rates get arbitrarily close to recent rates in expectation. Finally, we demonstrate how to apply our results to the training task in both supervised and reinforcement learning.
Authors: Harshit Varma, Dheeraj Nagaraj, Karthikeyan Shanmugam
Abstract: We introduce the Glauber Generative Model (GGM), a new class of discrete diffusion models, to obtain new samples from a distribution given samples from a discrete space. GGM deploys a discrete Markov chain called the heat bath dynamics (or the Glauber dynamics) to denoise a sequence of noisy tokens to a sample from a joint distribution of discrete tokens. Our novel conceptual framework provides an exact reduction of the task of learning the denoising Markov chain to solving a class of binary classification tasks. More specifically, the model learns to classify a given token in a noisy sequence as signal or noise. In contrast, prior works on discrete diffusion models either solve regression problems to learn importance ratios, or minimize loss functions given by variational approximations. We apply GGM to language modeling and image generation, where images are discretized using image tokenizers like VQGANs. We show that it outperforms existing discrete diffusion models in language generation, and demonstrates strong performance for image generation without using dataset-specific image tokenizers. We also show that our model is capable of performing well in zero-shot control settings like text and image infilling.
Authors: Dongjie Chen, Kartik Patwari, Zhengfeng Lai, Xiaoguang Zhu, Sen-ching Cheung, Chen-Nee Chuah
Abstract: Source-Free Domain Adaptation (SFDA) aims to adapt a pre-trained source model to a target domain using only unlabeled target data. Current SFDA methods face challenges in effectively leveraging pre-trained knowledge and exploiting target domain data. Multimodal Large Language Models (MLLMs) offer remarkable capabilities in understanding visual and textual information, but their applicability to SFDA poses challenges such as instruction-following failures, intensive computational demands, and difficulties in performance measurement prior to adaptation. To alleviate these issues, we propose $\textbf{Reliability-based Curriculum Learning (RCL)}$, a novel framework that integrates multiple MLLMs for knowledge exploitation via pseudo-labeling in SFDA. Our framework incorporates Reliable Knowledge Transfer, Self-correcting and MLLM-guided Knowledge Expansion, and Multi-hot Masking Refinement to progressively exploit unlabeled data in the target domain. RCL achieves state-of-the-art (SOTA) performance on multiple SFDA benchmarks, e.g., $\textbf{+9.4%}$ on DomainNet, demonstrating its effectiveness in enhancing adaptability and robustness without requiring access to source data. Our code is available at: https://github.com/Dong-Jie-Chen/RCL.
Authors: Shwai He, Daize Dong, Liang Ding, Ang Li
Abstract: Scaling large language models has driven remarkable advancements across various domains, yet the continual increase in model size presents significant challenges for real-world deployment. The Mixture of Experts (MoE) architecture offers a promising solution by dynamically selecting and activating only a subset of experts during inference, thus substantially reducing computational costs while preserving high performance. Despite these benefits, MoE introduces new inefficiencies, such as excessive parameters and communication overhead. In this work, we present a holistic study of compression techniques for Mixture of Experts to enhance both efficiency and scalability. While recent efforts have focused on Expert Trimming, which reduces the number of experts, these approaches still suffer from considerable communication and computational costs. To address this, we propose more aggressive strategies, such as Layer Drop, which removes entire MoE layers, and Block Drop, which eliminates transformer blocks. Surprisingly, these aggressive pruning techniques not only preserve model performance but also substantially improve computation and memory efficiency. Furthermore, beyond Expert Trimming, we also introduce Expert Slimming, which compresses individual experts to further boost performance and can be seamlessly integrated with Expert Trimming. Extensive experimental results demonstrate the effectiveness of our proposed methods-Layer Drop and Block Drop-along with the comprehensive recipe that integrates Expert Slimming and Expert Trimming, achieving a 6.05x speedup with 77.1% reduced memory usage while maintaining over 92% of performance on Mixtral-8x7B. Our code is released at https://github.com/CASE-Lab-UMD/Unified-MoE-Compression.
URLs: https://github.com/CASE-Lab-UMD/Unified-MoE-Compression.
Authors: Geeling Chau, Christopher Wang, Sabera Talukder, Vighnesh Subramaniam, Saraswati Soedarmadji, Yisong Yue, Boris Katz, Andrei Barbu
Abstract: We present a self-supervised framework that learns population-level codes for arbitrary ensembles of neural recordings at scale. We address key challenges in scaling models with neural time-series data, namely, sparse and variable electrode distribution across subjects and datasets. The Population Transformer (PopT) stacks on top of pretrained temporal embeddings and enhances downstream decoding by enabling learned aggregation of multiple spatially-sparse data channels. The pretrained PopT lowers the amount of data required for downstream decoding experiments, while increasing accuracy, even on held-out subjects and tasks. Compared to end-to-end methods, this approach is computationally lightweight, while achieving similar or better decoding performance. We further show how our framework is generalizable to multiple time-series embeddings and neural data modalities. Beyond decoding, we interpret the pretrained and fine-tuned PopT models to show how they can be used to extract neuroscience insights from large amounts of data. We release our code as well as a pretrained PopT to enable off-the-shelf improvements in multi-channel intracranial data decoding and interpretability. Code is available at https://github.com/czlwang/PopulationTransformer.
Authors: Md Atik Ahamed, Qiang Cheng
Abstract: Multivariate time series classification (TSC) is critical for various applications in fields such as healthcare and finance. While various approaches for TSC have been explored, important properties of time series, such as shift equivariance and inversion invariance, are largely underexplored by existing works. To fill this gap, we propose a novel multi-view approach to capture patterns with properties like shift equivariance. Our method integrates diverse features, including spectral, temporal, local, and global features, to obtain rich, complementary contexts for TSC. We use continuous wavelet transform to capture time-frequency features that remain consistent even when the input is shifted in time. These features are fused with temporal convolutional or multilayer perceptron features to provide complex local and global contextual information. We utilize the Mamba state space model for efficient and scalable sequence modeling and capturing long-range dependencies in time series. Moreover, we introduce a new scanning scheme for Mamba, called tango scanning, to effectively model sequence relationships and leverage inversion invariance, thereby enhancing our model's generalization and robustness. Experiments on two sets of benchmark datasets (10+20 datasets) demonstrate our approach's effectiveness, achieving average accuracy improvements of 4.01-6.45\% and 7.93\% respectively, over leading TSC models such as TimesNet and TSLANet.
Authors: Roman Belaire, Arunesh Sinha, Pradeep Varakantham
Abstract: Deep Reinforcement Learning (DRL) policies are highly susceptible to adversarial noise in observations, which poses significant risks in safety-critical scenarios. The challenge inherent to adversarial perturbations is that by altering the information observed by the agent, the state becomes only partially observable. Existing approaches address this by either enforcing consistent actions across nearby states or maximizing the worst-case value within adversarially perturbed observations. However, the former suffers from performance degradation when attacks succeed, while the latter tends to be overly conservative, leading to suboptimal performance in benign settings. We hypothesize that these limitations stem from their failing to account for partial observability directly. To this end, we introduce a novel objective called Adversarial Counterfactual Error (ACoE), defined on the beliefs about the true state and balancing value optimization with robustness. To make ACoE scalable in model-free settings, we propose the theoretically-grounded surrogate objective Cumulative-ACoE (C-ACoE). Our empirical evaluations on standard benchmarks (MuJoCo, Atari, and Highway) demonstrate that our method significantly outperforms current state-of-the-art approaches for addressing adversarial RL challenges, offering a promising direction for improving robustness in DRL under adversarial conditions. Our code is available at https://github.com/romanbelaire/acoe-robust-rl.
Authors: Marcin Przewi\k{e}\'zlikowski, Marcin Osial, Bartosz Zieli\'nski, Marek \'Smieja
Abstract: Collaborative self-supervised learning has recently become feasible in highly distributed environments by dividing the network layers between client devices and a central server. However, state-of-the-art methods, such as MocoSFL, are optimized for network division at the initial layers, which decreases the protection of the client data and increases communication overhead. In this paper, we demonstrate that splitting depth is crucial for maintaining privacy and communication efficiency in distributed training. We also show that MocoSFL suffers from a catastrophic quality deterioration for the minimal communication overhead. As a remedy, we introduce Momentum-Aligned contrastive Split Federated Learning (MonAcoSFL), which aligns online and momentum client models during training procedure. Consequently, we achieve state-of-the-art accuracy while significantly reducing the communication overhead, making MonAcoSFL more practical in real-world scenarios.
Authors: Naiqing Guan, Nick Koudas
Abstract: Efficient data annotation stands as a significant bottleneck in training contemporary machine learning models. The Programmatic Weak Supervision (PWS) pipeline presents a solution by utilizing multiple weak supervision sources to automatically label data, thereby expediting the annotation process. Given the varied contributions of these weak supervision sources to the accuracy of PWS, it is imperative to employ a robust and efficient metric for their evaluation. This is crucial not only for understanding the behavior and performance of the PWS pipeline but also for facilitating corrective measures. In our study, we introduce WeShap values as an evaluation metric, which quantifies the average contribution of weak supervision sources within a proxy PWS pipeline, leveraging the theoretical underpinnings of Shapley values. We demonstrate efficient computation of WeShap values using dynamic programming, achieving quadratic computational complexity relative to the number of weak supervision sources. Our experiments demonstrate the versatility of WeShap values across various applications, including the identification of beneficial or detrimental labeling functions, refinement of the PWS pipeline, and rectification of mislabeled data. Furthermore, WeShap values aid in comprehending the behavior of the PWS pipeline and scrutinizing specific instances of mislabeled data. Although initially derived from a specific proxy PWS pipeline, we empirically demonstrate the generalizability of WeShap values to other PWS pipeline configurations. Our findings indicate a noteworthy average improvement of 5.0 points in downstream model accuracy through the revision of the PWS pipeline compared to previous state-of-the-art methods, underscoring the efficacy of WeShap values in enhancing data quality for training machine learning models.
Authors: Weronika Ormaniec, Scott Sussex, Lars Lorch, Bernhard Sch\"olkopf, Andreas Krause
Abstract: Synthetic datasets generated by structural causal models (SCMs) are commonly used for benchmarking causal structure learning algorithms. However, the variances and pairwise correlations in SCM data tend to increase along the causal ordering. Several popular algorithms exploit these artifacts, possibly leading to conclusions that do not generalize to real-world settings. Existing metrics like $\operatorname{Var}$-sortability and $\operatorname{R^2}$-sortability quantify these patterns, but they do not provide tools to remedy them. To address this, we propose internally-standardized structural causal models (iSCMs), a modification of SCMs that introduces a standardization operation at each variable during the generative process. By construction, iSCMs are not $\operatorname{Var}$-sortable. We also find empirical evidence that they are mostly not $\operatorname{R^2}$-sortable for commonly-used graph families. Moreover, contrary to the post-hoc standardization of data generated by standard SCMs, we prove that linear iSCMs are less identifiable from prior knowledge on the weights and do not collapse to deterministic relationships in large systems, which may make iSCMs a useful model in causal inference beyond the benchmarking problem studied here. Our code is publicly available at: https://github.com/werkaaa/iscm.
Authors: Shengyuan Hu, Yiwei Fu, Zhiwei Steven Wu, Virginia Smith
Abstract: Machine unlearning is a promising approach to mitigate undesirable memorization of training data in ML models. However, in this work we show that existing approaches for unlearning in LLMs are surprisingly susceptible to a simple set of $\textit{benign relearning attacks}$. With access to only a small and potentially loosely related set of data, we find that we can ''jog'' the memory of unlearned models to reverse the effects of unlearning. For example, we show that relearning on public medical articles can lead an unlearned LLM to output harmful knowledge about bioweapons, and relearning general wiki information about the book series Harry Potter can force the model to output verbatim memorized text. We formalize this unlearning-relearning pipeline, explore the attack across three popular unlearning benchmarks, and discuss future directions and guidelines that result from our study. Our work indicates that current approximate unlearning methods simply suppress the model outputs and fail to robustly forget target knowledge in the LLMs.
Authors: Fu Feng, Yucheng Xie, Jing Wang, Xin Geng
Abstract: The growing complexity of model parameters underscores the significance of pre-trained models. However, deployment constraints often necessitate models of varying sizes, exposing limitations in the conventional pre-training and fine-tuning paradigm, particularly when target model sizes are incompatible with pre-trained ones. To address this challenge, we propose WAVE, a novel approach that reformulates variable-sized model initialization from a multi-task perspective, where initializing each model size is treated as a distinct task. WAVE employs shared, size-agnostic weight templates alongside size-specific weight scalers to achieve consistent initialization across various model sizes. These weight templates, constructed within the Learngene framework, integrate knowledge from pre-trained models through a distillation process constrained by Kronecker-based rules. Target models are then initialized by concatenating and weighting these templates, with adaptive connection rules established by lightweight weight scalers, whose parameters are learned from minimal training data. Extensive experiments demonstrate the efficiency of WAVE, achieving state-of-the-art performance in initializing models of various depth and width. The knowledge encapsulated in weight templates is also task-agnostic, allowing for seamless transfer across diverse downstream datasets. Code will be made available at https://github.com/fu-feng/WAVE.
Authors: Yi Ding, Chengxuan Tong, Shuailei Zhang, Muyun Jiang, Yong Li, Kevin Lim Jun Liang, Cuntai Guan
Abstract: Integrating prior knowledge of neurophysiology into neural network architecture enhances the performance of emotion decoding. While numerous techniques emphasize learning spatial and short-term temporal patterns, there has been limited emphasis on capturing the vital long-term contextual information associated with emotional cognitive processes. In order to address this discrepancy, we introduce a novel transformer model called emotion transformer (EmT). EmT is designed to excel in both generalized cross-subject EEG emotion classification and regression tasks. In EmT, EEG signals are transformed into a temporal graph format, creating a sequence of EEG feature graphs using a temporal graph construction module (TGC). A novel residual multi-view pyramid GCN module (RMPG) is then proposed to learn dynamic graph representations for each EEG feature graph within the series, and the learned representations of each graph are fused into one token. Furthermore, we design a temporal contextual transformer module (TCT) with two types of token mixers to learn the temporal contextual information. Finally, the task-specific output module (TSO) generates the desired outputs. Experiments on four publicly available datasets show that EmT achieves higher results than the baseline methods for both EEG emotion classification and regression tasks. The code is available at https://github.com/yi-ding-cs/EmT.
Authors: Yuxuan Wu, Ziyu Wang, Bhiksha Raj, Gus Xia
Abstract: We contribute an unsupervised method that effectively learns disentangled content and style representations from sequences of observations. Unlike most disentanglement algorithms that rely on domain-specific labels or knowledge, our method is based on the insight of domain-general statistical differences between content and style -- content varies more among different fragments within a sample but maintains an invariant vocabulary across data samples, whereas style remains relatively invariant within a sample but exhibits more significant variation across different samples. We integrate such inductive bias into an encoder-decoder architecture and name our method after V3 (variance-versus-invariance). Experimental results show that V3 generalizes across multiple domains and modalities, successfully learning disentangled content and style representations, such as pitch and timbre from music audio, digit and color from images of hand-written digits, and action and character appearance from simple animations. V3 demonstrates strong disentanglement performance compared to existing unsupervised methods, along with superior out-of-distribution generalization under few-shot adaptation compared to supervised counterparts. Lastly, symbolic-level interpretability emerges in the learned content codebook, forging a near one-to-one alignment between machine representation and human knowledge.
Authors: Matteo Gallici, Mattie Fellows, Benjamin Ellis, Bartomeu Pou, Ivan Masmitja, Jakob Nicolaus Foerster, Mario Martin
Abstract: Q-learning played a foundational role in the field reinforcement learning (RL). However, TD algorithms with off-policy data, such as Q-learning, or nonlinear function approximation like deep neural networks require several additional tricks to stabilise training, primarily a large replay buffer and target networks. Unfortunately, the delayed updating of frozen network parameters in the target network harms the sample efficiency and, similarly, the large replay buffer introduces memory and implementation overheads. In this paper, we investigate whether it is possible to accelerate and simplify off-policy TD training while maintaining its stability. Our key theoretical result demonstrates for the first time that regularisation techniques such as LayerNorm can yield provably convergent TD algorithms without the need for a target network or replay buffer, even with off-policy data. Empirically, we find that online, parallelised sampling enabled by vectorised environments stabilises training without the need for a large replay buffer. Motivated by these findings, we propose PQN, our simplified deep online Q-Learning algorithm. Surprisingly, this simple algorithm is competitive with more complex methods like: Rainbow in Atari, PPO-RNN in Craftax, QMix in Smax, and can be up to 50x faster than traditional DQN without sacrificing sample efficiency. In an era where PPO has become the go-to RL algorithm, PQN reestablishes off-policy Q-learning as a viable alternative.
Authors: Tongzhou Liao, Barnab\'as P\'oczos
Abstract: Graph Neural Networks (GNNs) have become important tools for machine learning on graph-structured data. In this paper, we explore the synergistic combination of graph encoding, graph rewiring, and graph attention, by introducing Graph Attention with Stochastic Structures (GRASS), a novel GNN architecture. GRASS utilizes relative random walk probabilities (RRWP) encoding and a novel decomposed variant (D-RRWP) to efficiently capture structural information. It rewires the input graph by superimposing a random regular graph to enhance long-range information propagation. It also employs a novel additive attention mechanism tailored for graph-structured data. Our empirical evaluations demonstrate that GRASS achieves state-of-the-art performance on multiple benchmark datasets, including a 20.3% reduction in mean absolute error on the ZINC dataset.
Authors: Arinbjorn Kolbeinsson, Kyle O'Brien, Tianjin Huang, Shanghua Gao, Shiwei Liu, Jonathan Richard Schwarz, Anurag Vaidya, Faisal Mahmood, Marinka Zitnik, Tianlong Chen, Thomas Hartvigsen
Abstract: Test-time interventions for language models can enhance factual accuracy, mitigate harmful outputs, and improve model efficiency without costly retraining. But despite a flood of new methods, different types of interventions are largely developing independently. In practice, multiple interventions must be applied sequentially to the same model, yet we lack standardized ways to study how interventions interact. We fill this gap by introducing composable interventions, a framework to study the effects of using multiple interventions on the same language models, featuring new metrics and a unified codebase. Using our framework, we conduct extensive experiments and compose popular methods from three emerging intervention categories -- Knowledge Editing, Model Compression, and Machine Unlearning. Our results from 310 different compositions uncover meaningful interactions: compression hinders editing and unlearning, composing interventions hinges on their order of application, and popular general-purpose metrics are inadequate for assessing composability. Taken together, our findings showcase clear gaps in composability, suggesting a need for new multi-objective interventions. All of our code is public: https://github.com/hartvigsen-group/composable-interventions.
URLs: https://github.com/hartvigsen-group/composable-interventions.
Authors: Bo Li, Wei Wang, Peng Ye
Abstract: Machine Learning has made remarkable progress in a wide range of fields. In many scenarios, learning is performed on datasets involving sensitive information, in which privacy protection is essential for learning algorithms. In this work, we study pure private learning in the agnostic model -- a framework reflecting the learning process in practice. We examine the number of users required under item-level (where each user contributes one example) and user-level (where each user contributes multiple examples) privacy and derive several improved upper bounds. For item-level privacy, our algorithm achieves a near optimal bound for general concept classes. We extend this to the user-level setting, rendering a tighter upper bound than the one proved by Ghazi et al. (2023). Lastly, we consider the problem of learning thresholds under user-level privacy and present an algorithm with a nearly tight user complexity.
Authors: Wei Huo, Changxin Liu, Kemi Ding, Karl Henrik Johansson, Ling Shi
Abstract: This paper investigates the use of the cubic-regularized Newton method within a federated learning framework while addressing two major concerns that commonly arise in federated learning: privacy leakage and communication bottleneck. We introduce a federated learning algorithm called Differentially Private Federated Cubic Regularized Newton (DP-FCRN). By leveraging second-order techniques, our algorithm achieves lower iteration complexity compared to first-order methods. We also incorporate noise perturbation during local computations to ensure privacy. Furthermore, we employ sparsification in uplink transmission, which not only reduces the communication costs but also amplifies the privacy guarantee. Specifically, this approach reduces the necessary noise intensity without compromising privacy protection. We analyze the convergence properties of our algorithm and establish the privacy guarantee. Finally, we validate the effectiveness of the proposed algorithm through experiments on a benchmark dataset.
Authors: Jiashu Tao, Reza Shokri
Abstract: Machine learning models can leak private information about their training data. The standard methods to measure this privacy risk, based on membership inference attacks (MIAs), only check if a given data point \textit{exactly} matches a training point, neglecting the potential of similar or partially overlapping memorized data revealing the same private information. To address this issue, we introduce the class of range membership inference attacks (RaMIAs), testing if the model was trained on any data in a specified range (defined based on the semantics of privacy). We formulate the RaMIAs game and design a principled statistical test for its composite hypotheses. We show that RaMIAs can capture privacy loss more accurately and comprehensively than MIAs on various types of data, such as tabular, image, and language. RaMIA paves the way for more comprehensive and meaningful privacy auditing of machine learning algorithms.
Authors: Sungmin Cha, Sungjun Cho, Dasol Hwang, Moontae Lee
Abstract: Large Language Models (LLMs) have demonstrated strong reasoning and memorization capabilities via pretraining on massive textual corpora. However, this poses risk of privacy and copyright violations, highlighting the need for efficient machine unlearning methods that remove sensitive data without retraining from scratch. While Gradient Ascent (GA) is commonly used to unlearn by reducing the likelihood of generating unwanted content, it leads to unstable optimization and catastrophic forgetting of retrained knowledge. We find that combining GA with low-rank adaptation results in poor trade-offs between computational cost and generative performance. To address these challenges, we propose two novel techniques for robust and efficient unlearning for LLMs. First, we introduce Inverted Hinge Loss, which suppresses unwanted tokens while maintaining fluency by boosting the probability of the next most likely token. Second, we develop a data-adaptive initialization for LoRA adapters via low-rank approximation weighted with relative Fisher information, thereby focusing updates on parameters critical for removing targeted knowledge. Experiments on the Training Data Extraction Challenge dataset using GPT-Neo models as well as on the TOFU benchmark with Phi-1.5B and Llama2-7B models demonstrate that our approach effectively removes sensitive information while maintaining reasoning and generative capabilities with minimal impact. Our implementation can be found in https://github.com/csm9493/efficient-llm-unlearning.
Authors: Soumitra Kundu, Gargi Panda, Saumik Bhattacharya, Aurobinda Routray, Rajlakshmi Guha
Abstract: Non-invasive and continuous blood pressure (BP) monitoring is essential for the early prevention of many cardiovascular diseases. Estimating arterial blood pressure (ABP) from photoplethysmography (PPG) has emerged as a promising solution. However, existing deep learning approaches for PPG-to-ABP reconstruction (PAR) encounter certain information loss, impacting the precision of the reconstructed signal. To overcome this limitation, we introduce an invertible neural network for PPG to ABP reconstruction (INN-PAR), which employs a series of invertible blocks to jointly learn the mapping between PPG and its gradient with the ABP signal and its gradient. INN-PAR efficiently captures both forward and inverse mappings simultaneously, thereby preventing information loss. By integrating signal gradients into the learning process, INN-PAR enhances the network's ability to capture essential high-frequency details, leading to more accurate signal reconstruction. Moreover, we propose a multi-scale convolution module (MSCM) within the invertible block, enabling the model to learn features across multiple scales effectively. We have experimented on two benchmark datasets, which show that INN-PAR significantly outperforms the state-of-the-art methods in both waveform reconstruction and BP measurement accuracy. Codes can be found at: https://github.com/soumitra1992/INNPAR-PPG2ABP.
Authors: Aldo Pacchiano
Abstract: Many works have developed no-regret algorithms for contextual bandits with function approximation, where the mean reward function over context-action pairs belongs to a function class. Although there are many approaches to this problem, one that has gained in importance is the use of algorithms based on the optimism principle such as optimistic least squares. It can be shown the regret of this algorithm scales as square root of the product of the eluder dimension (a statistical measure of the complexity of the function class), the logarithm of the function class size and the time horizon. Unfortunately, even if the variance of the measurement noise of the rewards at each time is changing and is very small, the regret of the optimistic least squares algorithm scales with square root of the time horizon. In this work we are the first to develop algorithms that satisfy regret bounds of scaling not with the square root of the time horizon, but the square root of the sum of the measurement variances in the setting of contextual bandits with function approximation when the variances are unknown. These bounds generalize existing techniques for deriving second order bounds in contextual linear problems.
Authors: Eleni D. Koronaki, Geremy Loachamin Suntaxi, Paris Papavasileiou, Dimitrios G. Giovanis, Martin Kathrein, Andreas G. Boudouvis, St\'ephane P. A. Bordas
Abstract: Important variables of processes are often categorical, i.e. names or labels representing, e.g. categories of inputs, or types of reactors or a sequence of steps. In this work, we use Natural Language Processing Models to derive embeddings of such inputs that represent their actual meaning, or reflect the "distances" between categories, i.e. how similar or dissimilar they are. This is a marked difference from the current standard practice of using binary, or one-hot encoding to replace categorical variables with sequences of ones and zeros. Combined with dimensionality reduction techniques, either linear such as Principal Component Analysis, or nonlinear such as Uniform Manifold Approximation and Projection, the proposed approach leads to a meaningful, low-dimensional feature space. The significance of obtaining meaningful embeddings is illustrated in the context of an industrial coating process for cutting tools that includes both numerical and categorical inputs. In this industrial process, subject matter expertise suggests that the categorical inputs are critical for determining the final outcome but this cannot be taken into account with the current state-of-the-art. The proposed approach enables feature importance which is a marked improvement compared to the current state-of-the-art in the encoding of categorical variables. The proposed approach is not limited to the case-study presented here and is suitable for applications with similar mix of categorical and numerical critical inputs.
Authors: Yoshiaki Kitazawa
Abstract: Density ratio estimation (DRE) is a core technique in machine learning used to capture relationships between two probability distributions. $f$-divergence loss functions, which are derived from variational representations of $f$-divergence, have become a standard choice in DRE for achieving cutting-edge performance. This study provides novel theoretical insights into DRE by deriving upper and lower bounds on the $L_p$ errors through $f$-divergence loss functions. These bounds apply to any estimator belonging to a class of Lipschitz continuous estimators, irrespective of the specific $f$-divergence loss function employed. The derived bounds are expressed as a product involving the data dimensionality and the expected value of the density ratio raised to the $p$-th power. Notably, the lower bound includes an exponential term that depends on the Kullback--Leibler (KL) divergence, revealing that the $L_p$ error increases significantly as the KL divergence grows when $p > 1$. This increase becomes even more pronounced as the value of $p$ grows. The theoretical insights are validated through numerical experiments.
Authors: Jinhui Bai, Lei Shi
Abstract: Inspired by the structure of spherical harmonics, we propose the truncated kernel stochastic gradient descent (T-kernel SGD) algorithm with a least-square loss function for spherical data fitting. T-kernel SGD introduces a novel regularization strategy by implementing stochastic gradient descent through a closed-form solution of the projection of the stochastic gradient in a low-dimensional subspace. In contrast to traditional kernel SGD, the regularization strategy implemented by T-kernel SGD is more effective in balancing bias and variance by dynamically adjusting the hypothesis space during iterations. The most significant advantage of the proposed algorithm is that it can achieve theoretically optimal convergence rates using a constant step size (independent of the sample size) while overcoming the inherent saturation problem of kernel SGD. Additionally, we leverage the structure of spherical polynomials to derive an equivalent T-kernel SGD, significantly reducing storage and computational costs compared to kernel SGD. Typically, T-kernel SGD requires only $\mathcal{O}(n^{1+\frac{d}{d-1}\epsilon})$ computational complexity and $\mathcal{O}(n^{\frac{d}{d-1}\epsilon})$ storage to achieve optimal rates for the d-dimensional sphere, where $0<\epsilon<\frac{1}{2}$ can be arbitrarily small if the optimal fitting or the underlying space possesses sufficient regularity. This regularity is determined by the smoothness parameter of the objective function and the decaying rate of the eigenvalues of the integral operator associated with the kernel function, both of which reflect the difficulty of the estimation problem. Our main results quantitatively characterize how this prior information influences the convergence of T-kernel SGD. The numerical experiments further validate the theoretical findings presented in this paper.
Authors: Alexander Levine, Peter Stone, Amy Zhang
Abstract: In order to train agents that can quickly adapt to new objectives or reward functions, efficient unsupervised representation learning in sequential decision-making environments can be important. Frameworks such as the Exogenous Block Markov Decision Process (Ex-BMDP) have been proposed to formalize this representation-learning problem (Efroni et al., 2022b). In the Ex-BMDP framework, the agent's high-dimensional observations of the environment have two latent factors: a controllable factor, which evolves deterministically within a small state space according to the agent's actions, and an exogenous factor, which represents time-correlated noise, and can be highly complex. The goal of the representation learning problem is to learn an encoder that maps from observations into the controllable latent space, as well as the dynamics of this space. Efroni et al. (2022b) has shown that this is possible with a sample complexity that depends only on the size of the controllable latent space, and not on the size of the noise factor. However, this prior work has focused on the episodic setting, where the controllable latent state resets to a specific start state after a finite horizon. By contrast, if the agent can only interact with the environment in a single continuous trajectory, prior works have not established sample-complexity bounds. We propose STEEL, the first provably sample-efficient algorithm for learning the controllable dynamics of an Ex-BMDP from a single trajectory, in the function approximation setting. STEEL has a sample complexity that depends only on the sizes of the controllable latent space and the encoder function class, and (at worst linearly) on the mixing time of the exogenous noise factor. We prove that STEEL is correct and sample-efficient, and demonstrate STEEL on two toy problems. Code is available at: https://github.com/midi-lab/steel.
Authors: Xavier Warin
Abstract: A new Kolmogorov-Arnold network (KAN) is proposed to approximate potentially irregular functions in high dimensions. We provide error bounds for this approximation, assuming that the Kolmogorov-Arnold expansion functions are sufficiently smooth. When the function is only continuous, we also provide universal approximation theorems. We show that it outperforms multilayer perceptrons in terms of accuracy and convergence speed. We also compare it with several proposed KAN networks: it outperforms all networks for irregular functions and achieves similar accuracy to the original spline-based KAN network for smooth functions. Finally, we compare some of the KAN networks in optimizing a French hydraulic valley.
Authors: Yichen Song, Jiaming Wang, Yunbo Wang, Xiaokang Yang
Abstract: In the realm of computational physics, an enduring topic is the numerical solutions to partial differential equations (PDEs). Recently, the attention of researchers has shifted towards Neural Operator methods, renowned for their capability to approximate ``operators'' -- mappings from functions to functions. Despite the universal approximation theorem within neural operators, ensuring error bounds often requires employing numerous Fourier layers. However, what about lightweight models? In response to this question, we introduce DimOL (Dimension-aware Operator Learning), drawing insights from dimensional analysis. To implement DimOL, we propose the ProdLayer, which can be seamlessly integrated into FNO-based and Transformer-based PDE solvers, enhancing their ability to handle sum-of-products structures inherent in many physical systems. Empirically, DimOL models achieve up to 48% performance gain within the PDE datasets. Furthermore, by analyzing Fourier components' weights, we can symbolically discern the physical significance of each term. This sheds light on the opaque nature of neural networks, unveiling underlying physical principles.
Authors: Jiale Chen, Dingling Yao, Adeel Pervez, Dan Alistarh, Francesco Locatello
Abstract: We propose Scalable Mechanistic Neural Network (S-MNN), an enhanced neural network framework designed for scientific machine learning applications involving long temporal sequences. By reformulating the original Mechanistic Neural Network (MNN) (Pervez et al., 2024), we reduce the computational time and space complexities from cubic and quadratic with respect to the sequence length, respectively, to linear. This significant improvement enables efficient modeling of long-term dynamics without sacrificing accuracy or interpretability. Extensive experiments demonstrate that S-MNN matches the original MNN in precision while substantially reducing computational resources. Consequently, S-MNN can drop-in replace the original MNN in applications, providing a practical and efficient tool for integrating mechanistic bottlenecks into neural network models of complex dynamical systems. Source code is available at https://github.com/IST-DASLab/ScalableMNN .
Authors: Christopher Fifty, Ronald G. Junkins, Dennis Duan, Aniketh Iyengar, Jerry W. Liu, Ehsan Amid, Sebastian Thrun, Christopher R\'e
Abstract: Vector Quantized Variational AutoEncoders (VQ-VAEs) are designed to compress a continuous input to a discrete latent space and reconstruct it with minimal distortion. They operate by maintaining a set of vectors -- often referred to as the codebook -- and quantizing each encoder output to the nearest vector in the codebook. However, as vector quantization is non-differentiable, the gradient to the encoder flows around the vector quantization layer rather than through it in a straight-through approximation. This approximation may be undesirable as all information from the vector quantization operation is lost. In this work, we propose a way to propagate gradients through the vector quantization layer of VQ-VAEs. We smoothly transform each encoder output into its corresponding codebook vector via a rotation and rescaling linear transformation that is treated as a constant during backpropagation. As a result, the relative magnitude and angle between encoder output and codebook vector becomes encoded into the gradient as it propagates through the vector quantization layer and back to the encoder. Across 11 different VQ-VAE training paradigms, we find this restructuring improves reconstruction metrics, codebook utilization, and quantization error. Our code is available at https://github.com/cfifty/rotation_trick.
Authors: Michael Lan, Philip Torr, Austin Meek, Ashkan Khakzar, David Krueger, Fazl Barez
Abstract: We investigate feature universality in large language models (LLMs), a research field that aims to understand how different models similarly represent concepts in the latent spaces of their intermediate layers. Demonstrating feature universality allows discoveries about latent representations to generalize across several models. However, comparing features across LLMs is challenging due to polysemanticity, in which individual neurons often correspond to multiple features rather than distinct ones, making it difficult to disentangle and match features across different models. To address this issue, we employ a method known as dictionary learning by using sparse autoencoders (SAEs) to transform LLM activations into more interpretable spaces spanned by neurons corresponding to individual features. After matching feature neurons across models via activation correlation, we apply representational space similarity metrics on SAE feature spaces across different LLMs. Our experiments reveal significant similarities in SAE feature spaces across various LLMs, providing new evidence for feature universality.
Authors: Arash Khajooeinejad, Fatemeh Sadat Masoumi, Masoumeh Chapariniya
Abstract: Hierarchical Reinforcement Learning (HRL) is well-suitedd for solving complex tasks by breaking them down into structured policies. However, HRL agents often struggle with efficient exploration and quick adaptation. To overcome these limitations, we propose integrating meta-learning into HRL to enable agents to learn and adapt hierarchical policies more effectively. Our method leverages meta-learning to facilitate rapid task adaptation using prior experience, while intrinsic motivation mechanisms drive efficient exploration by rewarding the discovery of novel states. Specifically, our agent employs a high-level policy to choose among multiple low-level policies within custom-designed grid environments. By incorporating gradient-based meta-learning with differentiable inner-loop updates, we optimize performance across a curriculum of progressively challenging tasks. Experimental results highlight that our metalearning-enhanced hierarchical agent significantly outperforms standard HRL approaches lacking meta-learning and intrinsic motivation. The agent demonstrates faster learning, greater cumulative rewards, and higher success rates in complex grid-based scenarios. These Findings underscore the effectiveness of combining meta-learning, curriculum learning, and intrinsic motivation to enhance the capability of HRL agents in tackling complex tasks.
Authors: Ge Li, Dong Tian, Hongyi Zhou, Xinkai Jiang, Rudolf Lioutikov, Gerhard Neumann
Abstract: This work introduces Transformer-based Off-Policy Episodic Reinforcement Learning (TOP-ERL), a novel algorithm that enables off-policy updates in the ERL framework. In ERL, policies predict entire action trajectories over multiple time steps instead of single actions at every time step. These trajectories are typically parameterized by trajectory generators such as Movement Primitives (MP), allowing for smooth and efficient exploration over long horizons while capturing high-level temporal correlations. However, ERL methods are often constrained to on-policy frameworks due to the difficulty of evaluating state-action values for entire action sequences, limiting their sample efficiency and preventing the use of more efficient off-policy architectures. TOP-ERL addresses this shortcoming by segmenting long action sequences and estimating the state-action values for each segment using a transformer-based critic architecture alongside an n-step return estimation. These contributions result in efficient and stable training that is reflected in the empirical results conducted on sophisticated robot learning environments. TOP-ERL significantly outperforms state-of-the-art RL methods. Thorough ablation studies additionally show the impact of key design choices on the model performance.
Authors: Yan Scholten, Stephan G\"unnemann
Abstract: Conformal prediction provides model-agnostic and distribution-free uncertainty quantification through prediction sets that are guaranteed to include the ground truth with any user-specified probability. Yet, conformal prediction is not reliable under poisoning attacks where adversaries manipulate both training and calibration data, which can significantly alter prediction sets in practice. As a solution, we propose reliable prediction sets (RPS): the first efficient method for constructing conformal prediction sets with provable reliability guarantees under poisoning. To ensure reliability under training poisoning, we introduce smoothed score functions that reliably aggregate predictions of classifiers trained on distinct partitions of the training data. To ensure reliability under calibration poisoning, we construct multiple prediction sets, each calibrated on distinct subsets of the calibration data. We then aggregate them into a majority prediction set, which includes a class only if it appears in a majority of the individual sets. Both proposed aggregations mitigate the influence of datapoints in the training and calibration data on the final prediction set. We experimentally validate our approach on image classification tasks, achieving strong reliability while maintaining utility and preserving coverage on clean data. Overall, our approach represents an important step towards more trustworthy uncertainty quantification in the presence of data poisoning.
Authors: Weronika Ormaniec, Felix Dangel, Sidak Pal Singh
Abstract: The Transformer architecture has inarguably revolutionized deep learning, overtaking classical architectures like multi-layer perceptrons (MLPs) and convolutional neural networks (CNNs). At its core, the attention block differs in form and functionality from most other architectural components in deep learning--to the extent that, in comparison to MLPs/CNNs, Transformers are more often accompanied by adaptive optimizers, layer normalization, learning rate warmup, etc. The root causes behind these outward manifestations and the precise mechanisms that govern them remain poorly understood. In this work, we bridge this gap by providing a fundamental understanding of what distinguishes the Transformer from the other architectures--grounded in a theoretical comparison of the (loss) Hessian. Concretely, for a single self-attention layer, (a) we first entirely derive the Transformer's Hessian and express it in matrix derivatives; (b) we then characterize it in terms of data, weight, and attention moment dependencies; and (c) while doing so further highlight the important structural differences to the Hessian of classical networks. Our results suggest that various common architectural and optimization choices in Transformers can be traced back to their highly non-linear dependencies on the data and weight matrices, which vary heterogeneously across parameters. Ultimately, our findings provide a deeper understanding of the Transformer's unique optimization landscape and the challenges it poses.
Authors: Zhaoyang Wang, Weilei He, Zhiyuan Liang, Xuchao Zhang, Chetan Bansal, Ying Wei, Weitong Zhang, Huaxiu Yao
Abstract: Recent self-rewarding large language models (LLM) have successfully applied LLM-as-a-Judge to iteratively improve the alignment performance without the need of human annotations for preference data. These methods commonly utilize the same LLM to act as both the policy model (which generates responses) and the reward model (which scores and ranks those responses). The ranked responses are then used as preference pairs to train the LLM via direct alignment technologies (e.g. DPO). However, it is noteworthy that throughout this process, there is no guarantee of accuracy in the rewarding and ranking, which is critical for ensuring accurate rewards and high-quality preference data. Empirical results from relatively small LLMs (e.g., 7B parameters) also indicate that improvements from self-rewarding may diminish after several iterations in certain situations, which we hypothesize is due to accumulated bias in the reward system. This bias can lead to unreliable preference data for training the LLM. To address this issue, we first formulate and analyze the generalized iterative preference fine-tuning framework for self-rewarding language model. We then introduce the regularization to this generalized framework to mitigate the overconfident preference labeling in the self-rewarding process. Based on this theoretical insight, we propose a Consistency Regularized sElf-rewarding lAnguage Model (CREAM) that leverages the consistency of rewards across different iterations to regularize the self-rewarding training, helping the model to learn from more reliable preference data. With this explicit regularization, our empirical results demonstrate the superiority of CREAM in improving both reward consistency and alignment performance. The code is publicly available at https://github.com/Raibows/CREAM.
Authors: Chenyu Wang, Masatoshi Uehara, Yichun He, Amy Wang, Tommaso Biancalani, Avantika Lal, Tommi Jaakkola, Sergey Levine, Hanchen Wang, Aviv Regev
Abstract: Recent studies have demonstrated the strong empirical performance of diffusion models on discrete sequences across domains from natural language to biological sequence generation. For example, in the protein inverse folding task, conditional diffusion models have achieved impressive results in generating natural-like sequences that fold back into the original structure. However, practical design tasks often require not only modeling a conditional distribution but also optimizing specific task objectives. For instance, we may prefer protein sequences with high stability. To address this, we consider the scenario where we have pre-trained discrete diffusion models that can generate natural-like sequences, as well as reward models that map sequences to task objectives. We then formulate the reward maximization problem within discrete diffusion models, analogous to reinforcement learning (RL), while minimizing the KL divergence against pretrained diffusion models to preserve naturalness. To solve this RL problem, we propose a novel algorithm, DRAKES, that enables direct backpropagation of rewards through entire trajectories generated by diffusion models, by making the originally non-differentiable trajectories differentiable using the Gumbel-Softmax trick. Our theoretical analysis indicates that our approach can generate sequences that are both natural-like and yield high rewards. While similar tasks have been recently explored in diffusion models for continuous domains, our work addresses unique algorithmic and theoretical challenges specific to discrete diffusion models, which arise from their foundation in continuous-time Markov chains rather than Brownian motion. Finally, we demonstrate the effectiveness of DRAKES in generating DNA and protein sequences that optimize enhancer activity and protein stability, respectively, important tasks for gene therapies and protein-based therapeutics.
Authors: Bruno Mlodozeniec, Runa Eschenhagen, Juhan Bae, Alexander Immer, David Krueger, Richard Turner
Abstract: Diffusion models have led to significant advancements in generative modelling. Yet their widespread adoption poses challenges regarding data attribution and interpretability. In this paper, we aim to help address such challenges in diffusion models by developing an influence functions framework. Influence function-based data attribution methods approximate how a model's output would have changed if some training data were removed. In supervised learning, this is usually used for predicting how the loss on a particular example would change. For diffusion models, we focus on predicting the change in the probability of generating a particular example via several proxy measurements. We show how to formulate influence functions for such quantities and how previously proposed methods can be interpreted as particular design choices in our framework. To ensure scalability of the Hessian computations in influence functions, we systematically develop K-FAC approximations based on generalised Gauss-Newton matrices specifically tailored to diffusion models. We recast previously proposed methods as specific design choices in our framework and show that our recommended method outperforms previous data attribution approaches on common evaluations, such as the Linear Data-modelling Score (LDS) or retraining without top influences, without the need for method-specific hyperparameter tuning.
Authors: Addison Kristanto Julistiono, Davoud Ataee Tarzanagh, Navid Azizan
Abstract: Attention mechanisms have revolutionized several domains of artificial intelligence, such as natural language processing and computer vision, by enabling models to selectively focus on relevant parts of the input data. While recent work has characterized the optimization dynamics of gradient descent (GD) in attention-based models and the structural properties of its preferred solutions, less is known about more general optimization algorithms such as mirror descent (MD). In this paper, we investigate the convergence properties and implicit biases of a family of MD algorithms tailored for softmax attention mechanisms, with the potential function chosen as the $p$-th power of the $\ell_p$-norm. Specifically, we show that these algorithms converge in direction to a generalized hard-margin SVM with an $\ell_p$-norm objective when applied to a classification problem using a softmax attention model. Notably, our theoretical results reveal that the convergence rate is comparable to that of traditional GD in simpler models, despite the highly nonlinear and nonconvex nature of the present problem. Additionally, we delve into the joint optimization dynamics of the key-query matrix and the decoder, establishing conditions under which this complex joint optimization converges to their respective hard-margin SVM solutions. Lastly, our numerical experiments on real data demonstrate that MD algorithms improve generalization over standard GD and excel in optimal token selection.
Authors: Minhyuk Seo, Hyunseo Koh, Jonghyun Choi
Abstract: The majority of online continual learning (CL) advocates single-epoch training and imposes restrictions on the size of replay memory. However, single-epoch training would incur a different amount of computations per CL algorithm, and the additional storage cost to store logit or model in addition to replay memory is largely ignored in calculating the storage budget. Arguing different computational and storage budgets hinder fair comparison among CL algorithms in practice, we propose to use floating point operations (FLOPs) and total memory size in Byte as a metric for computational and memory budgets, respectively, to compare and develop CL algorithms in the same 'total resource budget.' To improve a CL method in a limited total budget, we propose adaptive layer freezing that does not update the layers for less informative batches to reduce computational costs with a negligible loss of accuracy. In addition, we propose a memory retrieval method that allows the model to learn the same amount of knowledge as using random retrieval in fewer iterations. Empirical validations on the CIFAR-10/100, CLEAR-10/100, and ImageNet-1K datasets demonstrate that the proposed approach outperforms the state-of-the-art methods within the same total budget
Authors: Michael Noukhovitch, Shengyi Huang, Sophie Xhonneux, Arian Hosseini, Rishabh Agarwal, Aaron Courville
Abstract: The dominant paradigm for RLHF is online and on-policy RL: synchronously generating from the large language model (LLM) policy, labelling with a reward model, and learning using feedback on the LLM's own outputs. While performant, this paradigm is computationally inefficient. Inspired by classical deep RL literature, we propose separating generation and learning in RLHF. This enables asynchronous generation of new samples while simultaneously training on old samples, leading to faster training and more compute-optimal scaling. However, asynchronous training relies on an underexplored regime, online but off-policy RLHF: learning on samples from previous iterations of our model which give a worse training signal. We tackle the fundamental challenge in this regime: how much off-policyness can we tolerate for asynchronous training to speed up learning but maintain performance? Among several RLHF algorithms we test, online DPO is found to be most robust to off-policy data, and robustness increases with the scale of the policy model. We study further compute optimizations for asynchronous RLHF but find that they come at a performance cost, giving rise to a trade-off. We verify the scalability of asynchronous RLHF by training a general-purpose chatbot from LLaMA 3.1 8B on an instruction-following task ~40% faster than a synchronous run while matching final performance. Finally, we extend our results to math and reasoning to demonstrate asynchronous RL can finetune Rho 1B on GSM8k ~70% faster while matching synchronous accuracy.
Authors: Yaochen Hu, Mai Zeng, Ge Zhang, Pavel Rumiantsev, Liheng Ma, Yingxue Zhang, Mark Coates
Abstract: Graph Neural Networks (GNN) exhibit superior performance in graph representation learning, but their inference cost can be high, due to an aggregation operation that can require a memory fetch for a very large number of nodes. This inference cost is the major obstacle to deploying GNN models with \emph{online prediction} to reflect the potentially dynamic node features. To address this, we propose an approach to reduce the number of nodes that are included during aggregation. We achieve this through a sparse decomposition, learning to approximate node representations using a weighted sum of linearly transformed features of a carefully selected subset of nodes within the extended neighbourhood. The approach achieves linear complexity with respect to the average node degree and the number of layers in the graph neural network. We introduce an algorithm to compute the optimal parameters for the sparse decomposition, ensuring an accurate approximation of the original GNN model, and present effective strategies to reduce the training time and improve the learning process. We demonstrate via extensive experiments that our method outperforms other baselines designed for inference speedup, achieving significant accuracy gains with comparable inference times for both node classification and spatio-temporal forecasting tasks.
Authors: Guneet S. Dhillon, Xingjian Shi, Yee Whye Teh, Alex Smola
Abstract: Supervised fine-tuning (SFT) and alignment of large language models (LLMs) are key steps in providing a good user experience. However, the concept of an appropriate alignment is inherently application-dependent, and current methods often rely on heuristic choices to drive optimization. In this work, we formulate SFT and alignment as a constrained optimization problem: the LLM is fine-tuned on a task while being required to meet application-specific requirements, without resorting to heuristics. To solve this, we propose Lagrange Large Language Models (L3Ms), which employ logarithmic barriers to enforce the constraints. This approach allows for the customization of L3Ms across diverse applications while avoiding heuristic-driven processes. We experimentally demonstrate the versatility and efficacy of L3Ms in achieving tailored alignments for various applications.
Authors: Karthik Prakhya, Tolga Birdal, Alp Yurtsever
Abstract: Solving non-convex, NP-hard optimization problems is crucial for training machine learning models, including neural networks. However, non-convexity often leads to black-box machine learning models with unclear inner workings. While convex formulations have been used for verifying neural network robustness, their application to training neural networks remains less explored. In response to this challenge, we reformulate the problem of training infinite-width two-layer ReLU networks as a convex completely positive program in a finite-dimensional (lifted) space. Despite the convexity, solving this problem remains NP-hard due to the complete positivity constraint. To overcome this challenge, we introduce a semidefinite relaxation that can be solved in polynomial time. We then experimentally evaluate the tightness of this relaxation, demonstrating its competitive performance in test accuracy across a range of classification tasks.
Authors: Chenyu Wang, Sharut Gupta, Xinyi Zhang, Sana Tonekaboni, Stefanie Jegelka, Tommi Jaakkola, Caroline Uhler
Abstract: Multimodal representation learning seeks to relate and decompose information inherent in multiple modalities. By disentangling modality-specific information from information that is shared across modalities, we can improve interpretability and robustness and enable downstream tasks such as the generation of counterfactual outcomes. Separating the two types of information is challenging since they are often deeply entangled in many real-world applications. We propose Disentangled Self-Supervised Learning (DisentangledSSL), a novel self-supervised approach for learning disentangled representations. We present a comprehensive analysis of the optimality of each disentangled representation, particularly focusing on the scenario not covered in prior work where the so-called Minimum Necessary Information (MNI) point is not attainable. We demonstrate that DisentangledSSL successfully learns shared and modality-specific features on multiple synthetic and real-world datasets and consistently outperforms baselines on various downstream tasks, including prediction tasks for vision-language data, as well as molecule-phenotype retrieval tasks for biological data. The code is available at https://github.com/uhlerlab/DisentangledSSL.
Authors: Shi Dong, Xiaobei Niu, Rui Zhong, Zhifeng Wang, Mingzhang Zuo
Abstract: Accurate annotation of educational resources is crucial for effective personalized learning and resource recommendation in online education. However, fine-grained knowledge labels often overlap or share similarities, making it difficult for existing multi-label classification methods to differentiate them. The label distribution imbalance due to sparsity of human annotations further intensifies these challenges. To address these issues, this paper introduces RR2QC, a novel Retrieval Reranking method to multi-label Question Classification by leveraging label semantics and meta-label refinement. First, RR2QC improves the pre-training strategy by utilizing semantic relationships within and across label groups. Second, it introduces a class center learning task to align questions with label semantics during downstream training. Finally, this method decomposes labels into meta-labels and uses a meta-label classifier to rerank the retrieved label sequences. In doing so, RR2QC enhances the understanding and prediction capability of long-tail labels by learning from meta-labels that frequently appear in other labels. Additionally, a mathematical LLM is used to generate solutions for questions, extracting latent information to further refine the model's insights. Experimental results show that RR2QC outperforms existing methods in Precision@K and F1 scores across multiple educational datasets, demonstrating its effectiveness for online education applications. The code and datasets are available at https://github.com/78Erii/RR2QC.
Authors: Debmalya Mandal, Goran Radanovic
Abstract: We study the setting of \emph{performative reinforcement learning} where the deployed policy affects both the reward, and the transition of the underlying Markov decision process. Prior work~\parencite{MTR23} has addressed this problem under the tabular setting and established last-iterate convergence of repeated retraining with iteration complexity explicitly depending on the number of states. In this work, we generalize the results to \emph{linear Markov decision processes} which is the primary theoretical model of large-scale MDPs. The main challenge with linear MDP is that the regularized objective is no longer strongly convex and we want a bound that scales with the dimension of the features, rather than states which can be infinite. Our first result shows that repeatedly optimizing a regularized objective converges to a \emph{performatively stable policy}. In the absence of strong convexity, our analysis leverages a new recurrence relation that uses a specific linear combination of optimal dual solutions for proving convergence. We then tackle the finite sample setting where the learner has access to a set of trajectories drawn from the current policy. We consider a reparametrized version of the primal problem, and construct an empirical Lagrangian which is to be optimized from the samples. We show that, under a \emph{bounded coverage} condition, repeatedly solving a saddle point of this empirical Lagrangian converges to a performatively stable solution, and also construct a primal-dual algorithm that solves the empirical Lagrangian efficiently. Finally, we show several applications of the general framework of performative RL including multi-agent systems.
Authors: Bo Li, Wei Wang, Peng Ye
Abstract: Differential privacy (DP) is a formal notion that restricts the privacy leakage of an algorithm when running on sensitive data, in which privacy-utility trade-off is one of the central problems in private data analysis. In this work, we investigate the fundamental limits of differential privacy in online learning algorithms and present evidence that separates three types of constraints: no DP, pure DP, and approximate DP. We first describe a hypothesis class that is online learnable under approximate DP but not online learnable under pure DP under the adaptive adversarial setting. This indicates that approximate DP must be adopted when dealing with adaptive adversaries. We then prove that any private online learner must make an infinite number of mistakes for almost all hypothesis classes. This essentially generalizes previous results and shows a strong separation between private and non-private settings since a finite mistake bound is always attainable (as long as the class is online learnable) when there is no privacy requirement.
Authors: Yuewen Sun, Lingjing Kong, Guangyi Chen, Loka Li, Gongxu Luo, Zijian Li, Yixuan Zhang, Yujia Zheng, Mengyue Yang, Petar Stojanov, Eran Segal, Eric P. Xing, Kun Zhang
Abstract: Prevalent in biomedical applications (e.g., human phenotype research), multimodal datasets can provide valuable insights into the underlying physiological mechanisms. However, current machine learning (ML) models designed to analyze these datasets often lack interpretability and identifiability guarantees, which are essential for biomedical research. Recent advances in causal representation learning have shown promise in identifying interpretable latent causal variables with formal theoretical guarantees. Unfortunately, most current work on multimodal distributions either relies on restrictive parametric assumptions or yields only coarse identification results, limiting their applicability to biomedical research that favors a detailed understanding of the mechanisms. In this work, we aim to develop flexible identification conditions for multimodal data and principled methods to facilitate the understanding of biomedical datasets. Theoretically, we consider a nonparametric latent distribution (c.f., parametric assumptions in previous work) that allows for causal relationships across potentially different modalities. We establish identifiability guarantees for each latent component, extending the subspace identification results from previous work. Our key theoretical contribution is the structural sparsity of causal connections between modalities, which, as we will discuss, is natural for a large collection of biomedical systems. Empirically, we present a practical framework to instantiate our theoretical insights. We demonstrate the effectiveness of our approach through extensive experiments on both numerical and synthetic datasets. Results on a real-world human phenotype dataset are consistent with established biomedical research, validating our theoretical and methodological framework.
Authors: Riccardo Grazzi, Julien Siems, J\"org K. H. Franke, Arber Zela, Frank Hutter, Massimiliano Pontil
Abstract: Linear Recurrent Neural Networks (LRNNs) such as Mamba, RWKV, GLA, mLSTM, and DeltaNet have emerged as efficient alternatives to Transformers for long sequences. However, both Transformers and LRNNs struggle to perform state-tracking, which may impair performance in tasks such as code evaluation. In one forward pass, current architectures are unable to solve even parity, the simplest state-tracking task, which non-linear RNNs can handle effectively. Recently, Sarrof et al. (2024) demonstrated that the failure of LRNNs like Mamba to solve parity stems from restricting the value range of their diagonal state-transition matrices to $[0, 1]$ and that incorporating negative values can resolve this issue. We extend this result to non-diagonal LRNNs such as DeltaNet. We prove that finite precision LRNNs with state-transition matrices having only positive eigenvalues cannot solve parity, while non-triangular matrices are needed to count modulo $3$. Notably, we also prove that LRNNs can learn any regular language when their state-transition matrices are products of identity minus vector outer product matrices, each with eigenvalues in the range $[-1, 1]$. Our experiments confirm that extending the eigenvalue range of Mamba and DeltaNet to include negative values not only enables them to solve parity but consistently improves their performance on state-tracking tasks. We also show that state-tracking enabled LRNNs can be pretrained stably and efficiently at scale (1.3B parameters), achieving competitive performance on language modeling and showing promise on code and math tasks.
Authors: Ricardo Monta\~nana, Jos\'e A. G\'amez, Jos\'e M. Puerta
Abstract: We propose ODTE, a new ensemble that uses oblique decision trees as base classifiers. Additionally, we introduce STree, the base algorithm for growing oblique decision trees, which leverages support vector machines to define hyperplanes within the decision nodes. We embed a multiclass strategy -- one-vs-one or one-vs-rest -- at the decision nodes, allowing the model to directly handle non-binary classification tasks without the need to cluster instances into two groups, as is common in other approaches from the literature. In each decision node, only the best-performing model SVM -- the one that minimizes an impurity measure for the n-ary classification -- is retained, even if the learned SVM addresses a binary classification subtask. An extensive experimental study involving 49 datasets and various state-of-the-art algorithms for oblique decision tree ensembles has been conducted. Our results show that ODTE ranks consistently above its competitors, achieving significant performance gains when hyperparameters are carefully tuned. Moreover, the oblique decision trees learned through STree are more compact than those produced by other algorithms evaluated in our experiments.
Authors: Zhaoyi Li, Wenjie Mei, Ke Yu, Yang Bai, Shihua Li
Abstract: Learning models of dynamical systems with external inputs, which may be, for example, nonsmooth or piecewise, is crucial for studying complex phenomena and predicting future state evolution, which is essential for applications such as safety guarantees and decision-making. In this work, we introduce \emph{Input Concomitant Neural ODEs (ICODEs)}, which incorporate precise real-time input information into the learning process of the models, rather than treating the inputs as hidden parameters to be learned. The sufficient conditions to ensure the model's contraction property are provided to guarantee that system trajectories of the trained model converge to a fixed point, regardless of initial conditions across different training processes. We validate our method through experiments on several representative real dynamics: Single-link robot, DC-to-DC converter, motion dynamics of a rigid body, Rabinovich-Fabrikant equation, Glycolytic-glycogenolytic pathway model, and heat conduction equation. The experimental results demonstrate that our proposed ICODEs efficiently learn the ground truth systems, achieving superior prediction performance under both typical and atypical inputs. This work offers a valuable class of neural ODE models for understanding physical systems with explicit external input information, with potentially promising applications in fields such as physics and robotics. Our code is available online at https://github.com/EEE-ai59/ICODE.git.
Authors: Youssef Mansour, Reinhard Heckel
Abstract: We investigate biases in pretraining datasets for large language models (LLMs) through dataset classification experiments. Building on prior work demonstrating the existence of biases in popular computer vision datasets, we analyze popular open-source pretraining datasets for LLMs derived from CommonCrawl including C4, RefinedWeb, DolmaCC, RedPajama-V2, FineWeb, and DCLM-Baseline. Despite those datasets being obtained with similar curation steps, neural networks can classify surprisingly well which dataset a single text sequence belongs to, significantly better than a human can. This indicates that small differences in filtering and processing pipelines induce fingerprints evident in formatting, vocabulary, and content distributions. Those biases remain even when the text is rewritten with LLMs. Moreover, these biases propagate through training: Random sequences generated by models trained on those datasets can be classified well by a classifier trained on the original datasets. This can be leveraged to estimate the pretraining mixture proportions of the data sources.
Authors: Juan Rodriguez, Xiangru Jian, Siba Smarak Panigrahi, Tianyu Zhang, Aarash Feizi, Abhay Puri, Akshay Kalkunte, Fran\c{c}ois Savard, Ahmed Masry, Shravan Nayak, Rabiul Awal, Mahsa Massoud, Amirhossein Abaskohi, Zichao Li, Suyuchen Wang, Pierre-Andr\'e No\"el, Mats Leon Richter, Saverio Vadacchino, Shubham Agarwal, Sanket Biswas, Sara Shanian, Ying Zhang, Noah Bolger, Kurt MacDonald, Simon Fauvel, Sathwik Tejaswi, Srinivas Sunkara, Joao Monteiro, Krishnamurthy DJ Dvijotham, Torsten Scholak, Nicolas Chapados, Sepideh Kharagani, Sean Hughes, M. \"Ozsu, Siva Reddy, Marco Pedersoli, Yoshua Bengio, Christopher Pal, Issam Laradji, Spandana Gella, Perouz Taslakian, David Vazquez, Sai Rajeswar
Abstract: Multimodal AI has the potential to significantly enhance document-understanding tasks, such as processing receipts, understanding workflows, extracting data from documents, and summarizing reports. Code generation tasks that require long-structured outputs can also be enhanced by multimodality. Despite this, their use in commercial applications is often limited due to limited access to training data and restrictive licensing, which hinders open access. To address these limitations, we introduce BigDocs-7.5M, a high-quality, open-access dataset comprising 7.5 million multimodal documents across 30 tasks. We use an efficient data curation process to ensure our data is high-quality and license-permissive. Our process emphasizes accountability, responsibility, and transparency through filtering rules, traceable metadata, and careful content analysis. Additionally, we introduce BigDocs-Bench, a benchmark suite with 10 novel tasks where we create datasets that reflect real-world use cases involving reasoning over Graphical User Interfaces (GUI) and code generation from images. Our experiments show that training with BigDocs-Bench improves average performance up to 25.8% over closed-source GPT-4o in document reasoning and structured output tasks such as Screenshot2HTML or Image2Latex generation. Finally, human evaluations showed a preference for outputs from models trained on BigDocs over GPT-4o. This suggests that BigDocs can help both academics and the open-source community utilize and improve AI tools to enhance multimodal capabilities and document reasoning. The project is hosted at https://bigdocs.github.io .
Authors: Andrew Charbonneau, Katherine Deck, Tapio Schneider
Abstract: This paper presents a physics-constrained neural differential equation framework for parameterization, and employs it to model the time evolution of seasonal snow depth given hydrometeorological forcings. When trained on data from multiple SNOTEL sites, the parameterization predicts daily snow depth with under 9% median error and Nash Sutcliffe Efficiencies over 0.94 across a wide variety of snow climates. The parameterization also generalizes to new sites not seen during training, which is not often true for calibrated snow models. Requiring the parameterization to predict snow water equivalent in addition to snow depth only increases error to ~12%. The structure of the approach guarantees the satisfaction of physical constraints, enables these constraints during model training, and allows modeling at different temporal resolutions without additional retraining of the parameterization. These benefits hold potential in climate modeling, and could extend to other dynamical systems with physical constraints.
Authors: Dongwei Wang, Huanrui Yang
Abstract: Quantization is a critical step to enable efficient LLM serving under limited resource. However, previous research observes that certain weights in the LLM, known as outliers, are significantly sensitive to quantization noises. Existing quantization methods leave these outliers as floating points or higher precisions to retain performance, posting challenges on the efficient hardware deployment of the mixed-precision model. This work investigates an alternative way to tame the sensitive weights' impact on the quantization error, by reducing the loss Hessian trace with respect to outliers through an efficient fine-tuning process. We propose Noise Perturbation Fine-tuning (NPFT), which identifies outlier weights and add random weight perturbations on the outliers as the model going through a PEFT optimization. NPFT tames the sensitivity of outlier weights so that the quantized model performance can be improved without special treatment to the outliers. When applied to OPT and LLaMA models, our NPFT method achieves stable performance improvements for both uniform and non-uniform quantizers, while also offering better inference efficiency. Notably, the simplest RTN can achieve performance on par with GPTQ using our NPFT on LLaMA2-7B-4bits benchmark.
Authors: Yilei Zhao, Wentao Zhang, Tingran Yang, Yong Jiang, Fei Huang, Wei Yang Bryan Lim
Abstract: In financial trading, factor models are widely used to price assets and capture excess returns from mispricing. Recently, we have witnessed the rise of variational autoencoder-based latent factor models, which learn latent factors self-adaptively. While these models focus on modeling overall market conditions, they often fail to effectively capture the temporal patterns of individual stocks. Additionally, representing multiple factors as single values simplifies the model but limits its ability to capture complex relationships and dependencies. As a result, the learned factors are of low quality and lack diversity, reducing their effectiveness and robustness across different trading periods. To address these issues, we propose a Spatio-Temporal factOR Model based on dual vector quantized variational autoencoders, named STORM, which extracts features of stocks from temporal and spatial perspectives, then fuses and aligns these features at the fine-grained and semantic level, and represents the factors as multi-dimensional embeddings. The discrete codebooks cluster similar factor embeddings, ensuring orthogonality and diversity, which helps distinguish between different factors and enables factor selection in financial trading. To show the performance of the proposed factor model, we apply it to two downstream experiments: portfolio management on two stock datasets and individual trading tasks on six specific stocks. The extensive experiments demonstrate STORM's flexibility in adapting to downstream tasks and superior performance over baseline models.
Authors: Yeo Jin Jung, Claire Donnat
Abstract: We address the challenge of incorporating document-level metadata into topic modeling to improve topic mixture estimation. To overcome the computational complexity and lack of theoretical guarantees in existing Bayesian methods, we extend probabilistic latent semantic indexing (pLSI), a frequentist framework for topic modeling, by incorporating document-level covariates or known similarities between documents through a graph formalism. Modeling documents as nodes and edges denoting similarities, we propose a new estimator based on a fast graph-regularized iterative singular value decomposition (SVD) that encourages similar documents to share similar topic mixture proportions. We characterize the estimation error of our proposed method by deriving high-probability bounds and develop a specialized cross-validation method to optimize our regularization parameters. We validate our model through comprehensive experiments on synthetic datasets and three real-world corpora, demonstrating improved performance and faster inference compared to existing Bayesian methods.
Authors: Lily Hu
Abstract: Discussions of statistical criteria for fairness commonly convey the normative significance of calibration within groups by invoking what risk scores "mean." On the Same Meaning picture, group-calibrated scores "mean the same thing" (on average) across individuals from different groups and accordingly, guard against disparate treatment of individuals based on group membership. My contention is that calibration guarantees no such thing. Since concrete actual people belong to many groups, calibration cannot ensure the kind of consistent score interpretation that the Same Meaning picture implies matters for fairness, unless calibration is met within every group to which an individual belongs. Alas only perfect predictors may meet this bar. The Same Meaning picture thus commits a reference class fallacy by inferring from calibration within some group to the "meaning" or evidential value of an individual's score, because they are a member of that group. The reference class answer it presumes does not only lack justification; it is very likely wrong. I then show that the reference class problem besets not just calibration but other group statistical criteria that claim a close connection to fairness. Reflecting on the origins of this oversight opens a wider lens onto the predominant methodology in algorithmic fairness based on stylized cases.
Authors: Changchang Sun, Ren Wang, Yihua Zhang, Jinghan Jia, Jiancheng Liu, Gaowen Liu, Yan Yan, Sijia Liu
Abstract: Machine unlearning (MU), which seeks to erase the influence of specific unwanted data from already-trained models, is becoming increasingly vital in model editing, particularly to comply with evolving data regulations like the ``right to be forgotten''. Conventional approaches are predominantly model-based, typically requiring retraining or fine-tuning the model's weights to meet unlearning requirements. In this work, we approach the MU problem from a novel input perturbation-based perspective, where the model weights remain intact throughout the unlearning process. We demonstrate the existence of a proactive input-based unlearning strategy, referred to forget vector, which can be generated as an input-agnostic data perturbation and remains as effective as model-based approximate unlearning approaches. We also explore forget vector arithmetic, whereby multiple class-specific forget vectors are combined through simple operations (e.g., linear combinations) to generate new forget vectors for unseen unlearning tasks, such as forgetting arbitrary subsets across classes. Extensive experiments validate the effectiveness and adaptability of the forget vector, showcasing its competitive performance relative to state-of-the-art model-based methods. Codes are available at https://github.com/Changchangsun/Forget-Vector.
Authors: Kun Wu, Yinuo Zhao, Zhiyuan Xu, Zhengping Che, Chengxiang Yin, Chi Harold Liu, Feiferi Feng, Jian Tang
Abstract: Offline Reinforcement Learning (RL), which operates solely on static datasets without further interactions with the environment, provides an appealing alternative to learning a safe and promising control policy. The prevailing methods typically learn a conservative policy to mitigate the problem of Q-value overestimation, but it is prone to overdo it, leading to an overly conservative policy. Moreover, they optimize all samples equally with fixed constraints, lacking the nuanced ability to control conservative levels in a fine-grained manner. Consequently, this limitation results in a performance decline. To address the above two challenges in a united way, we propose a framework, Adaptive Conservative Level in Q-Learning (ACL-QL), which limits the Q-values in a mild range and enables adaptive control on the conservative level over each state-action pair, i.e., lifting the Q-values more for good transitions and less for bad transitions. We theoretically analyze the conditions under which the conservative level of the learned Q-function can be limited in a mild range and how to optimize each transition adaptively. Motivated by the theoretical analysis, we propose a novel algorithm, ACL-QL, which uses two learnable adaptive weight functions to control the conservative level over each transition. Subsequently, we design a monotonicity loss and surrogate losses to train the adaptive weight functions, Q-function, and policy network alternatively. We evaluate ACL-QL on the commonly used D4RL benchmark and conduct extensive ablation studies to illustrate the effectiveness and state-of-the-art performance compared to existing offline DRL baselines.
Authors: Minh-Duc Nguyen, Phuong Mai Dinh, Quang-Huy Nguyen, Long P. Hoang, Dung D. Le
Abstract: Expensive multi-objective optimization problems (EMOPs) are common in real-world scenarios where evaluating objective functions is costly and involves extensive computations or physical experiments. Current Pareto set learning methods for such problems often rely on surrogate models like Gaussian processes to approximate the objective functions. These surrogate models can become fragmented, resulting in numerous small uncertain regions between explored solutions. When using acquisition functions such as the Lower Confidence Bound (LCB), these uncertain regions can turn into pseudo-local optima, complicating the search for globally optimal solutions. To address these challenges, we propose a novel approach called SVH-PSL, which integrates Stein Variational Gradient Descent (SVGD) with Hypernetworks for efficient Pareto set learning. Our method addresses the issues of fragmented surrogate models and pseudo-local optima by collectively moving particles in a manner that smooths out the solution space. The particles interact with each other through a kernel function, which helps maintain diversity and encourages the exploration of underexplored regions. This kernel-based interaction prevents particles from clustering around pseudo-local optima and promotes convergence towards globally optimal solutions. Our approach aims to establish robust relationships between trade-off reference vectors and their corresponding true Pareto solutions, overcoming the limitations of existing methods. Through extensive experiments across both synthetic and real-world MOO benchmarks, we demonstrate that SVH-PSL significantly improves the quality of the learned Pareto set, offering a promising solution for expensive multi-objective optimization problems.
Authors: Tim Tsz-Kit Lau, Weijian Li, Chenwei Xu, Han Liu, Mladen Kolar
Abstract: An appropriate choice of batch sizes in large-scale model training is crucial, yet it involves an intrinsic yet inevitable dilemma: large-batch training improves training efficiency in terms of memory utilization, while generalization performance often deteriorates due to small amounts of gradient noise. Despite this dilemma, the common practice of choosing batch sizes in language model training often prioritizes training efficiency -- employing either constant large sizes with data parallelism or implementing batch size warmup schedules. However, such batch size schedule designs remain heuristic and often fail to adapt to training dynamics, presenting the challenge of designing adaptive batch size schedules. Given the abundance of available datasets and the data-hungry nature of language models, data parallelism has become an indispensable distributed training paradigm, enabling the use of larger batch sizes for gradient computation. However, vanilla data parallelism requires replicas of model parameters, gradients, and optimizer states at each worker, which prohibits training larger models with billions of parameters. To optimize memory usage, more advanced parallelism strategies must be employed. In this work, we propose general-purpose and theoretically principled adaptive batch size schedules compatible with data parallelism and model parallelism. We develop a practical implementation with PyTorch Fully Sharded Data Parallel, facilitating the pretraining of language models of different sizes. We empirically demonstrate that our proposed approaches outperform constant batch sizes and heuristic batch size warmup schedules in the pretraining of models in the Llama 2 family, with particular focus on smaller models with up to 3 billion parameters. We also establish theoretical convergence guarantees for such adaptive batch size schedules with Adam for general smooth nonconvex objectives.
Authors: Bangchen Yin, Jiaao Wang, Weitao Du, Pengbo Wang, Penghua Ying, Haojun Jia, Zisheng Zhang, Yuanqi Du, Carla P. Gomes, Graeme Henkelman, Chenru Duan, Hai Xiao
Abstract: Molecular dynamics simulations demand unprecedented accuracy and scalability to tackle grand challenges in energy materials, catalytic processes, and biomolecular design. To bridge this gap, we present AlphaNet, a local frame-based equivariant model that simultaneously advances computational efficiency and predictive precision for atomistic systems. By constructing equivariant local frames with learnable geometric transitions, AlphaNet encodes atomic environments with enhanced representational capacity, achieving state of the art accuracy in energy and force predictions. Extensive benchmarks spanning defected graphene, formate decomposition, inorganic bulks, and large-scale datasets (OC2M and Matbench Discovery) demonstrate its superior performance over existing neural network interatomic potentials while ensuring scalability across diverse system sizes. The synergy of accuracy, efficiency, and transferability positions AlphaNet as a transformative tool for simulating multiscale phenomena, from catalyst dynamics to energy storage interfaces, with direct implications for accelerating the discovery of functional materials and complex molecular systems.
Authors: Elena Albu, Shan Gao, Pieter Stijnen, Frank E. Rademakers, Bas C T van Bussel, Taya Collyer, Tina Hernandez-Boussard, Laure Wynants, Ben Van Calster
Abstract: Dynamic predictive modelling using electronic health record (EHR) data has gained significant attention in recent years. The reliability and trustworthiness of such models depend heavily on the quality of the underlying data, which is, in part, determined by the stages preceding the model development: data extraction from EHR systems and data preparation. In this article, we identified over forty challenges encountered during these stages and provide actionable recommendations for addressing them. These challenges are organized into four categories: cohort definition, outcome definition, feature engineering, and data cleaning. This comprehensive list serves as a practical guide for data extraction engineers and researchers, promoting best practices and improving the quality and real-world applicability of dynamic prediction models in clinical settings.
Authors: Qi Cheems Wang, Zehao Xiao, Yixiu Mao, Yun Qu, Jiayi Shen, Yiqin Lv, Xiangyang Ji
Abstract: Foundation models have revolutionized general-purpose problem-solving, offering rapid task adaptation through pretraining, meta-training, and finetuning. Recent crucial advances in these paradigms reveal the importance of challenging task prioritized sampling to enhance adaptation robustness under distribution shifts. However, ranking task difficulties over iteration as a preliminary step typically requires exhaustive task evaluation, which is practically unaffordable in computation and data-annotation. This study provides a novel perspective to illuminate the possibility of leveraging the dual importance of adaptation robustness and learning efficiency, particularly in scenarios where task evaluation is risky or costly, such as iterative agent-environment interactions for robotic policy evaluation or computationally intensive inference steps for finetuning foundation models. Firstly, we introduce Model Predictive Task Sampling (MPTS), a framework that bridges the task space and adaptation risk landscape, providing a theoretical foundation for robust active task sampling. MPTS employs a generative model to characterize the episodic optimization process and predicts task-specific adaptation risk via posterior inference. The resulting risk learner amortizes the costly evaluation of task adaptation performance and provably approximates task difficulty rankings. MPTS seamlessly integrates into zero-shot, few-shot, and supervised finetuning settings. Empirically, we conduct extensive experiments in pattern recognition using foundation models and sequential decision-making. Our results demonstrate that MPTS significantly enhances adaptation robustness for tail or out-of-distribution (OOD) tasks and improves learning efficiency compared to state-of-the-art (SOTA) methods. The code is available at the project site https://github.com/thu-rllab/MPTS.
Authors: Tian Jin, Ahmed Imtiaz Humayun, Utku Evci, Suvinay Subramanian, Amir Yazdanbakhsh, Dan Alistarh, Gintare Karolina Dziugaite
Abstract: Pruning eliminates unnecessary parameters in neural networks; it offers a promising solution to the growing computational demands of large language models (LLMs). While many focus on post-training pruning, sparse pre-training--which combines pruning and pre-training into a single phase--provides a simpler alternative. In this work, we present the first systematic exploration of optimal sparse pre-training configurations for LLMs through an examination of 80 unique pruning schedules across different sparsity levels and training durations. We find that initiating pruning at 25% of total training compute and concluding at 75% achieves near-optimal final evaluation loss. These findings provide valuable insights for efficient and effective sparse pre-training of LLMs. Furthermore, we propose a new scaling law that modifies the Chinchilla scaling law to use the average parameter count over pre-training. Through empirical and theoretical validation, we demonstrate that this modified scaling law accurately models evaluation loss for both sparsely and densely pre-trained LLMs, unifying scaling laws across pre-training paradigms. Our findings indicate that while sparse pre-training achieves the same final model quality as dense pre-training for equivalent compute budgets, it provides substantial benefits through reduced model size, enabling significant potential computational savings during inference.
Authors: Qian Chen, Lei Li, Qian Li, Jianghua Wu, Akang Wang, Ruoyu Sun, Xiaodong Luo, Tsung-Hui Chang, Qingjiang Shi
Abstract: A common characteristic in integer linear programs (ILPs) is symmetry, allowing variables to be permuted without altering the underlying problem structure. Recently, GNNs have emerged as a promising approach for solving ILPs. However, a significant challenge arises when applying GNNs to ILPs with symmetry: classic GNN architectures struggle to differentiate between symmetric variables, which limits their predictive accuracy. In this work, we investigate the properties of permutation equivariance and invariance in GNNs, particularly in relation to the inherent symmetry of ILP formulations. We reveal that the interaction between these two factors contributes to the difficulty of distinguishing between symmetric variables. To address this challenge, we explore the potential of feature augmentation and propose several guiding principles for constructing augmented features. Building on these principles, we develop an orbit-based augmentation scheme that first groups symmetric variables and then samples augmented features for each group from a discrete uniform distribution. Empirical results demonstrate that our proposed approach significantly enhances both training efficiency and predictive performance.
Authors: Jiazhen Chen, Sichao Fu, Zheng Ma, Mingbin Feng, Tony S. Wirjanto, Qinmu Peng
Abstract: Semi-supervised graph anomaly detection (GAD) has recently received increasing attention, which aims to distinguish anomalous patterns from graphs under the guidance of a moderate amount of labeled data and a large volume of unlabeled data. Although these proposed semi-supervised GAD methods have achieved great success, their superior performance will be seriously degraded when the provided labels are extremely limited due to some unpredictable factors. Besides, the existing methods primarily focus on anomaly detection in static graphs, and little effort was paid to consider the continuous evolution characteristic of graphs over time (dynamic graphs). To address these challenges, we propose a novel GAD framework (EL$^{2}$-DGAD) to tackle anomaly detection problem in dynamic graphs with extremely limited labels. Specifically, a transformer-based graph encoder model is designed to more effectively preserve evolving graph structures beyond the local neighborhood. Then, we incorporate an ego-context hypersphere classification loss to classify temporal interactions according to their structure and temporal neighborhoods while ensuring the normal samples are mapped compactly against anomalous data. Finally, the above loss is further augmented with an ego-context contrasting module which utilizes unlabeled data to enhance model generalization. Extensive experiments on four datasets and three label rates demonstrate the effectiveness of the proposed method in comparison to the existing GAD methods.
Authors: Ziqi Liu
Abstract: Long-term time series forecasting is essential in areas like finance and weather prediction. Besides traditional methods that operate in the time domain, many recent models transform time series data into the frequency domain to better capture complex patterns. However, these methods often use filtering techniques to remove certain frequency signals as noise, which may unintentionally discard important information and reduce prediction accuracy. To address this, we propose the Frequency Decomposition Mixture-of-Experts (FreqMoE) model, which dynamically decomposes time series data into frequency bands, each processed by a specialized expert. A gating mechanism adjusts the importance of each output of expert based on frequency characteristics, and the aggregated results are fed into a prediction module that iteratively refines the forecast using residual connections. Our experiments demonstrate that FreqMoE outperforms state-of-the-art models, achieving the best performance on 51 out of 70 metrics across all tested datasets, while significantly reducing the number of required parameters to under 50k, providing notable efficiency advantages. Code is available at: https://github.com/sunbus100/FreqMoE-main
Authors: Maximilian Muschalik, Fabian Fumagalli, Paolo Frazzetto, Janine Strotherm, Luca Hermes, Alessandro Sperduti, Eyke H\"ullermeier, Barbara Hammer
Abstract: Albeit the ubiquitous use of Graph Neural Networks (GNNs) in machine learning (ML) prediction tasks involving graph-structured data, their interpretability remains challenging. In explainable artificial intelligence (XAI), the Shapley Value (SV) is the predominant method to quantify contributions of individual features to a ML model's output. Addressing the limitations of SVs in complex prediction models, Shapley Interactions (SIs) extend the SV to groups of features. In this work, we explain single graph predictions of GNNs with SIs that quantify node contributions and interactions among multiple nodes. By exploiting the GNN architecture, we show that the structure of interactions in node embeddings are preserved for graph prediction. As a result, the exponential complexity of SIs depends only on the receptive fields, i.e. the message-passing ranges determined by the connectivity of the graph and the number of convolutional layers. Based on our theoretical results, we introduce GraphSHAP-IQ, an efficient approach to compute any-order SIs exactly. GraphSHAP-IQ is applicable to popular message passing techniques in conjunction with a linear global pooling and output layer. We showcase that GraphSHAP-IQ substantially reduces the exponential complexity of computing exact SIs on multiple benchmark datasets. Beyond exact computation, we evaluate GraphSHAP-IQ's approximation of SIs on popular GNN architectures and compare with existing baselines. Lastly, we visualize SIs of real-world water distribution networks and molecule structures using a SI-Graph.
Authors: Srinitish Srinivasan, Omkumar CU
Abstract: Graph representation learning has emerged as a cornerstone for tasks like node classification and link prediction, yet prevailing self-supervised learning (SSL) methods face challenges such as computational inefficiency, reliance on contrastive objectives, and representation collapse. Existing approaches often depend on feature reconstruction, negative sampling, or complex decoders, which introduce training overhead and hinder generalization. Further, current techniques which address such limitations fail to account for the contribution of node embeddings to a certain prediction in the absence of labeled nodes. To address these limitations, we propose a novel joint embedding predictive framework for graph SSL that eliminates contrastive objectives and negative sampling while preserving semantic and structural information. Additionally, we introduce a semantic-aware objective term that incorporates pseudo-labels derived from Gaussian Mixture Models (GMMs), enhancing node discriminability by evaluating latent feature contributions. Extensive experiments demonstrate that our framework outperforms state-of-the-art graph SSL methods across benchmarks, achieving superior performance without contrastive loss or complex decoders. Key innovations include (1) a non-contrastive, view-invariant joint embedding predictive architecture, (2) Leveraging single context and multiple targets relationship between subgraphs, and (3) GMM-based pseudo-label scoring to capture semantic contributions. This work advances graph SSL by offering a computationally efficient, collapse-resistant paradigm that bridges spatial and semantic graph features for downstream tasks. The code for our paper can be found at https://github.com/Deceptrax123/JPEB-GSSL
Authors: Lu Yi, Jie Peng, Yanping Zheng, Fengran Mo, Zhewei Wei, Yuhang Ye, Yue Zixuan, Zengfeng Huang
Abstract: Future link prediction is a fundamental challenge in various real-world dynamic systems. To address this, numerous temporal graph neural networks (temporal GNNs) and benchmark datasets have been developed. However, these datasets often feature excessive repeated edges and lack complex sequential dynamics, a key characteristic inherent in many real-world applications such as recommender systems and ``Who-To-Follow'' on social networks. This oversight has led existing methods to inadvertently downplay the importance of learning sequential dynamics, focusing primarily on predicting repeated edges. In this study, we demonstrate that existing methods, such as GraphMixer and DyGFormer, are inherently incapable of learning simple sequential dynamics, such as ``a user who has followed OpenAI and Anthropic is more likely to follow AI at Meta next.'' Motivated by this issue, we introduce the Temporal Graph Benchmark with Sequential Dynamics (TGB-Seq), a new benchmark carefully curated to minimize repeated edges, challenging models to learn sequential dynamics and generalize to unseen edges. TGB-Seq comprises large real-world datasets spanning diverse domains, including e-commerce interactions, movie ratings, business reviews, social networks, citation networks and web link networks. Benchmarking experiments reveal that current methods usually suffer significant performance degradation and incur substantial training costs on TGB-Seq, posing new challenges and opportunities for future research. TGB-Seq datasets, leaderboards, and example codes are available at https://tgb-seq.github.io/.
Authors: Liran Nochumsohn, Hedi Zisling, Omri Azencot
Abstract: Accurate forecasting of multivariate time series data is important in many engineering and scientific applications. Recent state-of-the-art works ignore the inter-relations between variates, using their model on each variate independently. This raises several research questions related to proper modeling of multivariate data. In this work, we propose to view multivariate forecasting as a multi-task learning problem, facilitating the analysis of forecasting by considering the angle between task gradients and their balance. To do so, we analyze linear models to characterize the behavior of tasks. Our analysis suggests that tasks can be defined by grouping similar variates together, which we achieve via a simple clustering that depends on correlation-based similarities. Moreover, to balance tasks, we scale gradients with respect to their prediction error. Then, each task is solved with a linear model within our MTLinear framework. We evaluate our approach on challenging benchmarks in comparison to strong baselines, and we show it obtains on-par or better results on multivariate forecasting problems. The implementation is available at: https://github.com/azencot-group/MTLinear
Authors: Tenglong Liu, Jianxiong Li, Yinan Zheng, Haoyi Niu, Yixing Lan, Xin Xu, Xianyuan Zhan
Abstract: Humans excel at reusing prior knowledge to address new challenges and developing skills while solving problems. This paradigm becomes increasingly popular in the development of autonomous agents, as it develops systems that can self-evolve in response to new challenges like human beings. However, previous methods suffer from limited training efficiency when expanding new skills and fail to fully leverage prior knowledge to facilitate new task learning. In this paper, we propose Parametric Skill Expansion and Composition (PSEC), a new framework designed to iteratively evolve the agents' capabilities and efficiently address new challenges by maintaining a manageable skill library. This library can progressively integrate skill primitives as plug-and-play Low-Rank Adaptation (LoRA) modules in parameter-efficient finetuning, facilitating efficient and flexible skill expansion. This structure also enables the direct skill compositions in parameter space by merging LoRA modules that encode different skills, leveraging shared information across skills to effectively program new skills. Based on this, we propose a context-aware module to dynamically activate different skills to collaboratively handle new tasks. Empowering diverse applications including multi-objective composition, dynamics shift, and continual policy shift, the results on D4RL, DSRL benchmarks, and the DeepMind Control Suite show that PSEC exhibits superior capacity to leverage prior knowledge to efficiently tackle new challenges, as well as expand its skill libraries to evolve the capabilities. Project website: https://ltlhuuu.github.io/PSEC/.
Authors: Junyi Guan, Abhijith Sharma, Chong Tian, Salem Lahlou
Abstract: Spiking Neural Networks (SNNs) are increasingly explored for their energy efficiency and robustness in real-world applications, yet their privacy risks remain largely unexamined. In this work, we investigate the susceptibility of SNNs to Membership Inference Attacks (MIAs) -- a major privacy threat where an adversary attempts to determine whether a given sample was part of the training dataset. While prior work suggests that SNNs may offer inherent robustness due to their discrete, event-driven nature, we find that its resilience diminishes as latency (T) increases. Furthermore, we introduce an input dropout strategy under black box setting, that significantly enhances membership inference in SNNs. Our findings challenge the assumption that SNNs are inherently more secure, and even though they are expected to be better, our results reveal that SNNs exhibit privacy vulnerabilities that are equally comparable to Artificial Neural Networks (ANNs). Our code is available at https://anonymous.4open.science/r/MIA_SNN-3610.
Authors: Adrien Aumon, Shuang Ni, Myriam Lizotte, Guy Wolf, Kevin R. Moon, Jake S. Rhodes
Abstract: Decades of research have produced robust methods for unsupervised data visualization, yet supervised visualization$\unicode{x2013}$where expert labels guide representations$\unicode{x2013}$remains underexplored, as most supervised approaches prioritize classification over visualization. Recently, RF-PHATE, a diffusion-based manifold learning method leveraging random forests and information geometry, marked significant progress in supervised visualization. However, its lack of an explicit mapping function limits scalability and prevents application to unseen data, posing challenges for large datasets and label-scarce scenarios. To overcome these limitations, we introduce Random Forest Autoencoders (RF-AE), a neural network-based framework for out-of-sample kernel extension that combines the flexibility of autoencoders with the supervised learning strengths of random forests and the geometry captured by RF-PHATE. RF-AE enables efficient out-of-sample supervised visualization and outperforms existing methods, including RF-PHATE's standard kernel extension, in both accuracy and interpretability. Additionally, RF-AE is robust to the choice of hyper-parameters and generalizes to any kernel-based dimensionality reduction method.
Authors: Jun Zhang, Jue Wang, Huan Li, Lidan Shou, Ke Chen, Yang You, Guiming Xie, Xuejian Gong, Kunlong Zhou
Abstract: Large Language Models (LLMs) have significantly advanced natural language processing with exceptional task generalization capabilities. Low-Rank Adaption (LoRA) offers a cost-effective fine-tuning solution, freezing the original model parameters and training only lightweight, low-rank adapter matrices. However, the memory footprint of LoRA is largely dominated by the original model parameters. To mitigate this, we propose LoRAM, a memory-efficient LoRA training scheme founded on the intuition that many neurons in over-parameterized LLMs have low training utility but are essential for inference. LoRAM presents a unique twist: it trains on a pruned (small) model to obtain pruned low-rank matrices, which are then recovered and utilized with the original (large) model for inference. Additionally, minimal-cost continual pre-training, performed by the model publishers in advance, aligns the knowledge discrepancy between pruned and original models. Our extensive experiments demonstrate the efficacy of LoRAM across various pruning strategies and downstream tasks. For a model with 70 billion parameters, LoRAM enables training on a GPU with only 20G HBM, replacing an A100-80G GPU for LoRA training and 15 GPUs for full fine-tuning. Specifically, QLoRAM implemented by structured pruning combined with 4-bit quantization, for LLaMA-3.1-70B (LLaMA-2-70B), reduces the parameter storage cost that dominates the memory usage in low-rank matrix training by 15.81$\times$ (16.95$\times$), while achieving dominant performance gains over both the original LLaMA-3.1-70B (LLaMA-2-70B) and LoRA-trained LLaMA-3.1-8B (LLaMA-2-13B). Code is available at https://github.com/junzhang-zj/LoRAM.
Authors: Shijin Duan, Yejia Liu, Gaowen Liu, Ramana Rao Kompella, Shaolei Ren, Xiaolin Xu
Abstract: Vector Symbolic Architecture (VSA) is emerging in machine learning due to its efficiency, but they are hindered by issues of hyperdimensionality and accuracy. As a promising mitigation, the Low-Dimensional Computing (LDC) method significantly reduces the vector dimension by ~100 times while maintaining accuracy, by employing a gradient-based optimization. Despite its potential, LDC optimization for VSA is still underexplored. Our investigation into vector updates underscores the importance of stable, adaptive dynamics in LDC training. We also reveal the overlooked yet critical roles of batch normalization (BN) and knowledge distillation (KD) in standard approaches. Besides the accuracy boost, BN does not add computational overhead during inference, and KD significantly enhances inference confidence. Through extensive experiments and ablation studies across multiple benchmarks, we provide a thorough evaluation of our approach and extend the interpretability of binary neural network optimization similar to LDC, previously unaddressed in BNN literature.
Authors: Shailik Sarkar (Virginia Tech), Raquib Bin Yousuf (Virginia Tech), Linhan Wang (Virginia Tech), Brian Mayer (Virginia Tech), Thomas Mortier (World Forest ID), Victor Deklerck (World Forest ID), Jakub Truszkowski (World Forest ID), John C. Simeone (Simeone Consulting LLC), Marigold Norman (World Forest ID), Jade Saunders (World Forest ID), Chang-Tien Lu (Virginia Tech), Naren Ramakrishnan (Virginia Tech)
Abstract: Illegal logging poses a significant threat to global biodiversity, climate stability, and depresses international prices for legal wood harvesting and responsible forest products trade, affecting livelihoods and communities across the globe. Stable isotope ratio analysis (SIRA) is rapidly becoming an important tool for determining the harvest location of traded, organic, products. The spatial pattern in stable isotope ratio values depends on factors such as atmospheric and environmental conditions and can thus be used for geographic origin identification. We present here the results of a deployed machine learning pipeline where we leverage both isotope values and atmospheric variables to determine timber harvest location. Additionally, the pipeline incorporates uncertainty estimation to facilitate the interpretation of harvest location determination for analysts. We present our experiments on a collection of oak (Quercus spp.) tree samples from its global range. Our pipeline outperforms comparable state-of-the-art models determining geographic harvest origin of commercially traded wood products, and has been used by European enforcement agencies to identify harvest location misrepresentation. We also identify opportunities for further advancement of our framework and how it can be generalized to help identify the origin of falsely labeled organic products throughout the supply chain.
Authors: Raquib Bin Yousuf, Hoang Anh Just, Shengzhe Xu, Brian Mayer, Victor Deklerck, Jakub Truszkowski, John C. Simeone, Jade Saunders, Chang-Tien Lu, Ruoxi Jia, Naren Ramakrishnan
Abstract: Determining and verifying product provenance remains a critical challenge in global supply chains, particularly as geopolitical conflicts and shifting borders create new incentives for misrepresentation of commodities, such as hiding the origin of illegally harvested timber or agriculture grown on illegally cleared land. Stable Isotope Ratio Analysis (SIRA), combined with Gaussian process regression-based isoscapes, has emerged as a powerful tool for geographic origin verification. However, the effectiveness of these models is often constrained by data scarcity and suboptimal dataset selection. In this work, we introduce a novel data valuation framework designed to enhance the selection and utilization of training data for machine learning models applied in SIRA. By prioritizing high-informative samples, our approach improves model robustness and predictive accuracy across diverse datasets and geographies. We validate our methodology with extensive experiments, demonstrating its potential to significantly enhance provenance verification, mitigate fraudulent trade practices, and strengthen regulatory enforcement of global supply chains.
Authors: Botian Wang, Yawen Ouyang, Yaohui Li, Yiqun Wang, Haorui Cui, Jianbing Zhang, Xiaonan Wang, Wei-Ying Ma, Hao Zhou
Abstract: Deep learning methods for material property prediction have been widely explored to advance materials discovery. However, the prevailing pre-train then fine-tune paradigm often fails to address the inherent diversity and disparity of material tasks. To overcome these challenges, we introduce MoMa, a Modular framework for Materials that first trains specialized modules across a wide range of tasks and then adaptively composes synergistic modules tailored to each downstream scenario. Evaluation across 17 datasets demonstrates the superiority of MoMa, with a substantial 14% average improvement over the strongest baseline. Few-shot and continual learning experiments further highlight MoMa's potential for real-world applications. Pioneering a new paradigm of modular material learning, MoMa will be open-sourced to foster broader community collaboration.
Authors: Arshia Kermani, Ehsan Zeraatkar, Habib Irani
Abstract: The increasing computational demands of transformer models in time series classification necessitate effective optimization strategies for energy-efficient deployment. Our study presents a systematic investigation of optimization techniques, focusing on structured pruning and quantization methods for transformer architectures. Through extensive experimentation on three distinct datasets (RefrigerationDevices, ElectricDevices, and PLAID), we quantitatively evaluate model performance and energy efficiency across different transformer configurations. Our experimental results demonstrate that static quantization reduces energy consumption by 29.14% while maintaining classification performance, and L1 pruning achieves a 63% improvement in inference speed with minimal accuracy degradation. Our findings provide valuable insights into the effectiveness of optimization strategies for transformer-based time series classification, establishing a foundation for efficient model deployment in resource-constrained environments.
Authors: Andrei Chernov, Oleg Novitskij
Abstract: Recent studies have shown that reducing symmetries in neural networks enhances linear mode connectivity between networks without requiring parameter space alignment, leading to improved performance in linearly interpolated neural networks. However, in practical applications, neural network interpolation is rarely used; instead, ensembles of networks are more common. In this paper, we empirically investigate the impact of reducing symmetries on the performance of deep ensembles and Mixture of Experts (MoE) across five datasets. Additionally, to explore deeper linear mode connectivity, we introduce the Mixture of Interpolated Experts (MoIE). Our results show that deep ensembles built on asymmetric neural networks achieve significantly better performance as ensemble size increases compared to their symmetric counterparts. In contrast, our experiments do not provide conclusive evidence on whether reducing symmetries affects both MoE and MoIE architectures.
Authors: Yuhan Chen, Yihong Luo, Yifan Song, Pengwen Dai, Jing Tang, Xiaochun Cao
Abstract: Despite extensive research efforts focused on OOD detection on images, OOD detection on nodes in graph learning remains underexplored. The dependence among graph nodes hinders the trivial adaptation of existing approaches on images that assume inputs to be i.i.d. sampled, since many unique features and challenges specific to graphs are not considered, such as the heterophily issue. Recently, GNNSafe, which considers node dependence, adapted energy-based detection to the graph domain with state-of-the-art performance, however, it has two serious issues: 1) it derives node energy from classification logits without specifically tailored training for modeling data distribution, making it less effective at recognizing OOD data; 2) it highly relies on energy propagation, which is based on homophily assumption and will cause significant performance degradation on heterophilic graphs, where the node tends to have dissimilar distribution with its neighbors. To address the above issues, we suggest training EBMs by MLE to enhance data distribution modeling and remove energy propagation to overcome the heterophily issues. However, training EBMs via MLE requires performing MCMC sampling on both node feature and node neighbors, which is challenging due to the node interdependence and discrete graph topology. To tackle the sampling challenge, we introduce DeGEM, which decomposes the learning process into two parts: a graph encoder that leverages topology information for node representations and an energy head that operates in latent space. Extensive experiments validate that DeGEM, without OOD exposure during training, surpasses previous state-of-the-art methods, achieving an average AUROC improvement of 6.71% on homophilic graphs and 20.29% on heterophilic graphs, and even outperform methods trained with OOD exposure. Our code is available at: https://github.com/draym28/DeGEM.
Authors: Jacob Fein-Ashley, Rajgopal Kannan, Viktor Prasanna
Abstract: Conventional self-attention mechanisms exhibit quadratic complexity in sequence length, making them challenging to scale for long inputs. We present FFTNet, an adaptive spectral filtering framework that uses the Fast Fourier Transform (FFT) to achieve global token mixing in \(\mathcal{O}(n\log n)\) time. By mapping inputs into the frequency domain, FFTNet exploits orthogonality and energy preservation-guaranteed by Parseval's theorem-to efficiently model long-range dependencies. Our main theoretical contributions include 1) An adaptive spectral filter that highlights salient frequency components, 2) A hybrid scheme combining local windowing with a global FFT branch, 3) Nonlinear feature transformations applied in both the frequency and token domains. Experiments on Long Range Arena and ImageNet validate our theoretical insights and demonstrate superior performance over fixed Fourier-based and standard attention models.
Authors: Maher Hanut, Jonathan Kadmon
Abstract: Training deep neural networks typically relies on backpropagating high dimensional error signals a computationally intensive process with little evidence supporting its implementation in the brain. However, since most tasks involve low-dimensional outputs, we propose that low-dimensional error signals may suffice for effective learning. To test this hypothesis, we introduce a novel local learning rule based on Feedback Alignment that leverages indirect, low-dimensional error feedback to train large networks. Our method decouples the backward pass from the forward pass, enabling precise control over error signal dimensionality while maintaining high-dimensional representations. We begin with a detailed theoretical derivation for linear networks, which forms the foundation of our learning framework, and extend our approach to nonlinear, convolutional, and transformer architectures. Remarkably, we demonstrate that even minimal error dimensionality on the order of the task dimensionality can achieve performance matching that of traditional backpropagation. Furthermore, our rule enables efficient training of convolutional networks, which have previously been resistant to Feedback Alignment methods, with minimal error. This breakthrough not only paves the way toward more biologically accurate models of learning but also challenges the conventional reliance on high-dimensional gradient signals in neural network training. Our findings suggest that low-dimensional error signals can be as effective as high-dimensional ones, prompting a reevaluation of gradient-based learning in high-dimensional systems. Ultimately, our work offers a fresh perspective on neural network optimization and contributes to understanding learning mechanisms in both artificial and biological systems.
Authors: Bo Yang
Abstract: Large language models memorize sensitive data from their pretraining corpora. In this work, we propose CE-U (Cross Entropy Unlearning), a loss function for unlearning. CE-U addresses fundamental limitations of gradient ascent approaches that suffer from vanishing gradients when model confidence is high and exploding gradients when confidence is low. We also unify standard cross entropy learning and unlearning into a single framework. On the TOFU benchmark for unlearning, CE-U achieves state-of-the-art results on LLaMA2-7B models without using an extra oracle model or additional positive samples. Our analysis reveals that the problematic gradient ascent component also exists in reinforcement learning algorithms like DPO and GRPO. This suggests that applying CE-U approach to reinforcement learning could be promising to improve stability and convergence.
Authors: Jianqi Yan (The University of Hong Kong), Alex P. Leung (The University of Hong Kong), Zhiyuan Pei (Macau University of Science and Technology), David C. Y. Hui (Chungnam National University), Sangin Kim (Chungnam National University)
Abstract: This work introduces a novel deep learning-based approach for gravitational wave anomaly detection, aiming to overcome the limitations of traditional matched filtering techniques in identifying unknown waveform gravitational wave signals. We introduce a modified convolutional neural network architecture inspired by ResNet that leverages residual blocks to extract high-dimensional features, effectively capturing subtle differences between background noise and gravitational wave signals. This network architecture learns a high-dimensional projection while preserving discrepancies with the original input, facilitating precise identification of gravitational wave signals. In our experiments, we implement an innovative data augmentation strategy that generates new data by computing the arithmetic mean of multiple signal samples while retaining the key features of the original signals. In the NSF HDR A3D3: Detecting Anomalous Gravitational Wave Signals competition, it is honorable for us (group name: easonyan123) to get to the first place at the end with our model achieving a true negative rate (TNR) of 0.9708 during development/validation phase and 0.9832 on an unseen challenge dataset during final/testing phase, the highest among all competitors. These results demonstrate that our method not only achieves excellent generalization performance but also maintains robust adaptability in addressing the complex uncertainties inherent in gravitational wave anomaly detection.
Authors: Houyi Li, Wenzheng Zheng, Jingcheng Hu, Qiufeng Wang, Hanshan Zhang, Zili Wang, Shijie Xuyang, Yuantao Fan, Shuigeng Zhou, Xiangyu Zhang, Daxin Jiang
Abstract: The impressive capabilities of Large Language Models (LLMs) across diverse tasks are now well-established, yet their effective deployment necessitates careful hyperparameter optimization. Through extensive empirical studies involving grid searches across diverse configurations, we discover universal scaling laws governing these hyperparameters: optimal learning rate follows a power-law relationship with both model parameters and data sizes, while optimal batch size scales primarily with data sizes. Our analysis reveals a convex optimization landscape for hyperparameters under fixed models and data size conditions. This convexity implies an optimal hyperparameter plateau. We contribute a universal, plug-and-play optimal hyperparameter tool for the community. Its estimated values on the test set are merely 0.09% away from the globally optimal LLM performance found via an exhaustive search. These laws demonstrate remarkable robustness across variations in model sparsity, training data distribution, and model shape. To our best known, this is the first work that unifies different model shapes and structures, such as Mixture-of-Experts models and dense transformers, as well as establishes optimal hyperparameter scaling laws across diverse data distributions. This exhaustive optimization process demands substantial computational resources, utilizing nearly one million NVIDIA H800 GPU hours to train 3,700 LLMs of varying sizes and hyperparameters from scratch and consuming approximately 100 trillion tokens in total. To facilitate reproducibility and further research, we will progressively release all loss measurements and model checkpoints through our designated repository https://step-law.github.io/
Authors: Jacob Gildenblat, Jens Pahnke
Abstract: We present Preserving Clusters and Correlations (PCC), a novel dimensionality reduction (DR) method a novel dimensionality reduction (DR) method that achieves state-of-the-art global structure (GS) preservation while maintaining competitive local structure (LS) preservation. It optimizes two objectives: a GS preservation objective that preserves an approximation of Pearson and Spearman correlations between high- and low-dimensional distances, and an LS preservation objective that ensures clusters in the high-dimensional data are separable in the low-dimensional data. PCC has a state-of-the-art ability to preserve the GS while having competitive LS preservation. In addition, we show the correlation objective can be combined with UMAP to significantly improve its GS preservation with minimal degradation of the LS. We quantitatively benchmark PCC against existing methods and demonstrate its utility in medical imaging, and show PCC is a competitive DR technique that demonstrates superior GS preservation in our benchmarks.
Authors: Chao Ning, Wanshui Gan, Weihao Xuan, Naoto Yokoya
Abstract: Pre-trained encoders are widely employed in dense prediction tasks for their capability to effectively extract visual features from images. The decoder subsequently processes these features to generate pixel-level predictions. However, due to structural differences and variations in input data, only encoders benefit from pre-learned representations from vision benchmarks such as image classification and self-supervised learning, while decoders are typically trained from scratch. In this paper, we introduce $\times$Net, which facilitates a "pre-trained encoder $\times$ pre-trained decoder" collaboration through three innovative designs. $\times$Net enables the direct utilization of pre-trained models within the decoder, integrating pre-learned representations into the decoding process to enhance performance in dense prediction tasks. By simply coupling the pre-trained encoder and pre-trained decoder, $\times$Net distinguishes itself as a highly promising approach. Remarkably, it achieves this without relying on decoding-specific structures or task-specific algorithms. Despite its streamlined design, $\times$Net outperforms advanced methods in tasks such as monocular depth estimation and semantic segmentation, achieving state-of-the-art performance particularly in monocular depth estimation. and semantic segmentation, achieving state-of-the-art results, especially in monocular depth estimation. embedding algorithms. Despite its streamlined design, $\times$Net outperforms advanced methods in tasks such as monocular depth estimation and semantic segmentation, achieving state-of-the-art performance particularly in monocular depth estimation.
Authors: Martin Kuhn, Joscha Gr\"uger, Tobias Geyer, Ralph Bergmann
Abstract: The rapid progress in modern medicine presents physicians with complex challenges when planning patient treatment. Techniques from the field of Predictive Business Process Monitoring, like Next-activity-prediction (NAP) can be used as a promising technique to support physicians in treatment planning, by proposing a possible next treatment step. Existing patient data, often in the form of electronic health records, can be analyzed to recommend the next suitable step in the treatment process. However, the use of patient data poses many challenges due to its knowledge-intensive character, high variability and scarcity of medical data. To overcome these challenges, this article examines the use of the knowledge encoded in taxonomies to improve and explain the prediction of the next activity in the treatment process. This study proposes the TS4NAP approach, which uses medical taxonomies (ICD-10-CM and ICD-10-PCS) in combination with graph matching to assess the similarities of medical codes to predict the next treatment step. The effectiveness of the proposed approach will be evaluated using event logs that are derived from the MIMIC-IV dataset. The results highlight the potential of using domain-specific knowledge held in taxonomies to improve the prediction of the next activity, and thus can improve treatment planning and decision-making by making the predictions more explainable.
Authors: Sara Alkhalifa
Abstract: The research presents a machine learning (ML) classifier designed to differentiate between schizophrenia patients and healthy controls by utilising features extracted from electroencephalogram (EEG) data, specifically focusing on event-related potentials (ERPs) and certain demographic variables. The dataset comprises data from 81 participants, encompassing 32 healthy controls and 49 schizophrenia patients, all sourced from an online dataset. After preprocessing the dataset, our ML model achieved an accuracy of 99.930%. This performance outperforms earlier research, including those that used deep learning methods. Additionally, an analysis was conducted to assess individual features' contribution to improving classification accuracy. This involved systematically excluding specific features from the original dataset one at a time, and another technique involved an iterative process of removing features based on their entropy scores incrementally. The impact of these removals on model performance was evaluated to identify the most informative features.
Authors: Kwanyoung Kim, Byeongsu Sim
Abstract: Diffusion models have shown impressive results in generating high-quality conditional samples using guidance techniques such as Classifier-Free Guidance (CFG). However, existing methods often require additional training or neural function evaluations (NFEs), making them incompatible with guidance-distilled models. Also, they rely on heuristic approaches that need identifying target layers. In this work, we propose a novel and efficient method, termed PLADIS, which boosts pre-trained models (U-Net/Transformer) by leveraging sparse attention. Specifically, we extrapolate query-key correlations using softmax and its sparse counterpart in the cross-attention layer during inference, without requiring extra training or NFEs. By leveraging the noise robustness of sparse attention, our PLADIS unleashes the latent potential of text-to-image diffusion models, enabling them to excel in areas where they once struggled with newfound effectiveness. It integrates seamlessly with guidance techniques, including guidance-distilled models. Extensive experiments show notable improvements in text alignment and human preference, offering a highly efficient and universally applicable solution. See Our project page : https://cubeyoung.github.io/pladis-proejct/
Authors: Jannis O. Luebsen, Annika Eichler
Abstract: This paper addresses the integration of additional information sources into a Bayesian optimization framework while ensuring that safety constraints are satisfied. The interdependencies between these information sources are modeled using an unknown correlation matrix. We explore how uniform error bounds must be adjusted to maintain constraint satisfaction throughout the optimization process, considering both Bayesian and frequentist statistical perspectives. This is achieved by appropriately scaling the error bounds based on a confidence interval that can be estimated from the data. Furthermore, the efficacy of the proposed approach is demonstrated through experiments on two benchmark functions and a controller parameter optimization problem. Our results highlight a significant improvement in sample efficiency, demonstrating the methods suitability for optimizing expensive-to-evaluate functions.
Authors: Yihong Huang, Chen Chu, Fan Zhang, Fei Chen, Yu Lin, Ruiduan Li, Zhihao Li
Abstract: Deep models in industrial applications rely on thousands of features for accurate predictions, such as deep recommendation systems. While new features are introduced to capture evolving user behavior, outdated or redundant features often remain, significantly increasing storage and computational costs. To address this issue, feature selection methods are widely adopted to identify and remove less important features. However, existing approaches face two major challenges: (1) they often require complex hyperparameter (Hp) tuning, making them difficult to employ in practice, and (2) they fail to produce well-separated feature importance scores, which complicates straightforward feature removal. Moreover, the impact of removing unimportant features can only be evaluated through retraining the model, a time-consuming and resource-intensive process that severely hinders efficient feature selection. To solve these challenges, we propose a novel feature selection approach, ShuffleGate. In particular, it shuffles all feature values across instances simultaneously and uses a gating mechanism that allows the model to dynamically learn the weights for combining the original and shuffled inputs. Notably, it can generate well-separated feature importance scores and estimate the performance without retraining the model, while introducing only a single Hp. Experiments on four public datasets show that our approach outperforms state-of-the-art methods in feature selection for model retraining. Moreover, it has been successfully integrated into the daily iteration of Bilibili's search models across various scenarios, where it significantly reduces feature set size (up to 60%+) and computational resource usage (up to 20%+), while maintaining comparable performance.
Authors: Andreas Kofler, Luca Calatroni, Christoph Kolbitsch, Kostas Papafitsoros
Abstract: We propose an unrolled algorithm approach for learning spatially adaptive parameter maps in the framework of convolutional synthesis-based $\ell_1$ regularization. More precisely, we consider a family of pre-trained convolutional filters and estimate deeply parametrized spatially varying parameters applied to the sparse feature maps by means of unrolling a FISTA algorithm to solve the underlying sparse estimation problem. The proposed approach is evaluated for image reconstruction of low-field MRI and compared to spatially adaptive and non-adaptive analysis-type procedures relying on Total Variation regularization and to a well-established model-based deep learning approach. We show that the proposed approach produces visually and quantitatively comparable results with the latter approaches and at the same time remains highly interpretable. In particular, the inferred parameter maps quantify the local contribution of each filter in the reconstruction, which provides valuable insight into the algorithm mechanism and could potentially be used to discard unsuited filters.
Authors: Yuanmin Huang, Mi Zhang, Zhaoxiang Wang, Wenxuan Li, Min Yang
Abstract: Time series classification (TSC) is a cornerstone of modern web applications, powering tasks such as financial data analysis, network traffic monitoring, and user behavior analysis. In recent years, deep neural networks (DNNs) have greatly enhanced the performance of TSC models in these critical domains. However, DNNs are vulnerable to backdoor attacks, where attackers can covertly implant triggers into models to induce malicious outcomes. Existing backdoor attacks targeting DNN-based TSC models remain elementary. In particular, early methods borrow trigger designs from computer vision, which are ineffective for time series data. More recent approaches utilize generative models for trigger generation, but at the cost of significant computational complexity. In this work, we analyze the limitations of existing attacks and introduce an enhanced method, FreqBack. Drawing inspiration from the fact that DNN models inherently capture frequency domain features in time series data, we identify that improper perturbations in the frequency domain are the root cause of ineffective attacks. To address this, we propose to generate triggers both effectively and efficiently, guided by frequency analysis. FreqBack exhibits substantial performance across five models and eight datasets, achieving an impressive attack success rate of over 90%, while maintaining less than a 3% drop in model accuracy on clean data.
Authors: Tomek Diederen, Nicola Zamboni
Abstract: We present a framework for modeling complex, high-dimensional distributions on convex polytopes by leveraging recent advances in discrete and continuous normalizing flows on Riemannian manifolds. We show that any full-dimensional polytope is homeomorphic to a unit ball, and our approach harnesses flows defined on the ball, mapping them back to the original polytope. Furthermore, we introduce a strategy to construct flows when only the vertex representation of a polytope is available, employing maximum entropy barycentric coordinates and Aitchison geometry. Our experiments take inspiration from applications in metabolic flux analysis and demonstrate that our methods achieve competitive density estimation, sampling accuracy, as well as fast training and inference times.
Authors: Bowen Wang, Matteo Zecchin, Osvaldo Simeone
Abstract: Online conformal prediction enables the runtime calibration of a pre-trained artificial intelligence model using feedback on its performance. Calibration is achieved through set predictions that are updated via online rules so as to ensure long-term coverage guarantees. While recent research has demonstrated the benefits of incorporating prior knowledge into the calibration process, this has come at the cost of replacing coverage guarantees with less tangible regret guarantees based on the quantile loss. This work introduces intermittent mirror online conformal prediction (IM-OCP), a novel runtime calibration framework that integrates prior knowledge, while maintaining long-term coverage and achieving sub-linear regret. IM-OCP features closed-form updates with minimal memory complexity, and is designed to operate under potentially intermittent feedback.
Authors: Jordan S. Ellenberg, Cristofero S. Fraser-Taliente, Thomas R. Harvey, Karan Srivastava, Andrew V. Sutherland
Abstract: We present a new implementation of the LLM-driven genetic algorithm {\it funsearch}, whose aim is to generate examples of interest to mathematicians and which has already had some success in problems in extremal combinatorics. Our implementation is designed to be useful in practice for working mathematicians; it does not require expertise in machine learning or access to high-performance computing resources. Applying {\it funsearch} to a new problem involves modifying a small segment of Python code and selecting a large language model (LLM) from one of many third-party providers. We benchmarked our implementation on three different problems, obtaining metrics that may inform applications of {\it funsearch} to new problems. Our results demonstrate that {\it funsearch} successfully learns in a variety of combinatorial and number-theoretic settings, and in some contexts learns principles that generalize beyond the problem originally trained on.
Authors: I. Bentley, J. Tedder, M. Gebran, A. Paul
Abstract: Sixteen new physics informed machine learning models have been trained on binding energy residuals from modern mass models that leverage shape parameters and other physical features. The models have been trained on a subset of AME 2012 data and have been verified with a subset of the AME 2020 data. Among the machine learning approaches tested in this work, the preferred approach is the least squares boosted ensemble of trees which appears to have a superior ability to both interpolate and extrapolate binding energy residuals. The machine learning models for four mass models created from the ensemble of trees approach have been combined to create a composite model called the Four Model Tree Ensemble (FMTE). The FMTE model predicts binding energy values from AME 2020 with a standard deviation of 76 keV and a mean deviation of 34 keV for all nuclei with N > 7 and Z > 7. A comparison with new mass measurements for 33 isotopes not included in AME 2012 or AME 2020 indicates that the FMTE performs better than all mass models that were tested.
Authors: James Ferlez, Mahmoud Elnaggar, Yasser Shoukry, Cody Fleming
Abstract: In this paper, we develop a novel closed-form Control Barrier Function (CBF) and associated controller shield for the Kinematic Bicycle Model (KBM) with respect to obstacle avoidance. The proposed CBF and shield -- designed by an algorithm we call ShieldNN -- provide two crucial advantages over existing methodologies. First, ShieldNN considers steering and velocity constraints directly with the non-affine KBM dynamics; this is in contrast to more general methods, which typically consider only affine dynamics and do not guarantee invariance properties under control constraints. Second, ShieldNN provides a closed-form set of safe controls for each state unlike more general methods, which typically rely on optimization algorithms to generate a single instantaneous for each state. Together, these advantages make ShieldNN uniquely suited as an efficient Multi-Obstacle Safe Actions (i.e. multiple-barrier-function shielding) during training time of a Reinforcement Learning (RL) enabled Neural Network controller. We show via experiments that ShieldNN dramatically increases the completion rate of RL training episodes in the presence of multiple obstacles, thus establishing the value of ShieldNN in training RL-based controllers.
Authors: Andrea Della Vecchia, Ernesto De Vito, Jaouad Mourtada, Lorenzo Rosasco
Abstract: We investigate an extension of classical empirical risk minimization, where the hypothesis space consists of a random subspace within a given Hilbert space. Specifically, we examine the Nystr\"om method where the subspaces are defined by a random subset of the data. This approach recovers Nystr\"om approximations used in kernel methods as a specific case. Using random subspaces naturally leads to computational advantages, but a key question is whether it compromises the learning accuracy. Recently, the tradeoffs between statistics and computation have been explored for the square loss and self-concordant losses, such as the logistic loss. In this paper, we extend these analyses to general convex Lipschitz losses, which may lack smoothness, such as the hinge loss used in support vector machines. Our main results show the existence of various scenarios where computational gains can be achieved without sacrificing learning performance. When specialized to smooth loss functions, our analysis recovers most previous results. Moreover, it allows to consider classification problems and translate the surrogate risk bounds into classification error bounds. Indeed, this gives the opportunity to compare the effect of Nystr\"om approximations when combined with different loss functions such as the hinge or the square loss.
Authors: Rahul Singh, Liyuan Xu, Arthur Gretton
Abstract: We propose simple nonparametric estimators for mediated and time-varying dose response curves based on kernel ridge regression. By embedding Pearl's mediation formula and Robins' g-formula with kernels, we allow treatments, mediators, and covariates to be continuous in general spaces, and also allow for nonlinear treatment-confounder feedback. Our key innovation is a reproducing kernel Hilbert space technique called sequential kernel embedding, which we use to construct simple estimators that account for complex feedback. Our estimators preserve the generality of classic identification while also achieving nonasymptotic uniform rates. In nonlinear simulations with many covariates, we demonstrate strong performance. We estimate mediated and time-varying dose response curves of the US Job Corps, and clean data that may serve as a benchmark in future work. We extend our results to mediated and time-varying treatment effects and counterfactual distributions, verifying semiparametric efficiency and weak convergence.
Authors: Archismita Dalal, Mohsen Bagherimehrab, Barry C. Sanders
Abstract: A popular machine-learning model for regression tasks, including stock-market prediction, weather forecasting and real-estate pricing, is the classical support vector regression (SVR). However, a practically realisable quantum SVR remains to be formulated. We devise annealing-based algorithms, namely simulated and quantum-classical hybrid, for training two SVR models and compare their empirical performances against the SVR implementation of Python's scikit-learn package for facial-landmark detection (FLD), a particular use case for SVR. Our method is to derive a quadratic-unconstrained-binary formulation for the optimisation problem used for training a SVR model and solve this problem using annealing. Using D-Wave's hybrid solver, we construct a quantum-assisted SVR model, thereby demonstrating a slight advantage over classical models regarding FLD accuracy. Furthermore, we observe that annealing-based SVR models predict landmarks with lower variances compared to the SVR models trained by gradient-based methods. Our work is a proof-of-concept example for applying quantum-assisted SVR to a supervised-learning task with a small training dataset.
Authors: Junqi Tang, Matthias Ehrhardt, Carola-Bibiane Sch\"onlieb
Abstract: In this work we propose a stochastic primal-dual three-operator splitting algorithm (TOS-SPDHG) for solving a class of convex three-composite optimization problems. Our proposed scheme is a direct three-operator splitting extension of the SPDHG algorithm [Chambolle et al. 2018]. We provide theoretical convergence analysis showing ergodic $O(1/K)$ convergence rate, and demonstrate the effectiveness of our approach in imaging inverse problems. Moreover, we further propose TOS-SPDHG-RED and TOS-SPDHG-eRED which utilizes the regularization-by-denoising (RED) framework to leverage pretrained deep denoising networks as priors.
Authors: Liyuan Hu, Mengbing Li, Chengchun Shi, Zhenke Wu, Piotr Fryzlewicz
Abstract: This paper studies reinforcement learning (RL) in doubly inhomogeneous environments under temporal non-stationarity and subject heterogeneity. In a number of applications, it is commonplace to encounter datasets generated by system dynamics that may change over time and population, challenging high-quality sequential decision making. Nonetheless, most existing RL solutions require either temporal stationarity or subject homogeneity, which would result in sub-optimal policies if both assumptions were violated. To address both challenges simultaneously, we propose an original algorithm to determine the ``best data chunks" that display similar dynamics over time and across individuals for policy learning, which alternates between most recent change point detection and cluster identification. Our method is general, and works with a wide range of clustering and change point detection algorithms. It is multiply robust in the sense that it takes multiple initial estimators as input and only requires one of them to be consistent. Moreover, by borrowing information over time and population, it allows us to detect weaker signals and has better convergence properties when compared to applying the clustering algorithm per time or the change point detection algorithm per subject. Empirically, we demonstrate the usefulness of our method through extensive simulations and a real data application.
Authors: Florian Kalinke, Zolt\'an Szab\'o
Abstract: Kernel techniques are among the most popular and powerful approaches of data science. Among the key features that make kernels ubiquitous are (i) the number of domains they have been designed for, (ii) the Hilbert structure of the function class associated to kernels facilitating their statistical analysis, and (iii) their ability to represent probability distributions without loss of information. These properties give rise to the immense success of Hilbert-Schmidt independence criterion (HSIC) which is able to capture joint independence of random variables under mild conditions, and permits closed-form estimators with quadratic computational complexity (w.r.t. the sample size). In order to alleviate the quadratic computational bottleneck in large-scale applications, multiple HSIC approximations have been proposed, however these estimators are restricted to $M=2$ random variables, do not extend naturally to the $M\ge 2$ case, and lack theoretical guarantees. In this work, we propose an alternative Nystr\"om-based HSIC estimator which handles the $M\ge 2$ case, prove its consistency, and demonstrate its applicability in multiple contexts, including synthetic examples, dependency testing of media annotations, and causal discovery.
Authors: Huanjing Yue, Cong Cao, Lei Liao, Jingyu Yang
Abstract: In recent years, raw video denoising has garnered increased attention due to the consistency with the imaging process and well-studied noise modeling in the raw domain. However, two problems still hinder the denoising performance. Firstly, there is no large dataset with realistic motions for supervised raw video denoising, as capturing noisy and clean frames for real dynamic scenes is difficult. To address this, we propose recapturing existing high-resolution videos displayed on a 4K screen with high-low ISO settings to construct noisy-clean paired frames. In this way, we construct a video denoising dataset (named as ReCRVD) with 120 groups of noisy-clean videos, whose ISO values ranging from 1600 to 25600. Secondly, while non-local temporal-spatial attention is beneficial for denoising, it often leads to heavy computation costs. We propose an efficient raw video denoising transformer network (RViDeformer) that explores both short and long-distance correlations. Specifically, we propose multi-branch spatial and temporal attention modules, which explore the patch correlations from local window, local low-resolution window, global downsampled window, and neighbor-involved window, and then they are fused together. We employ reparameterization to reduce computation costs. Our network is trained in both supervised and unsupervised manners, achieving the best performance compared with state-of-the-art methods. Additionally, the model trained with our proposed dataset (ReCRVD) outperforms the model trained with previous benchmark dataset (CRVD) when evaluated on the real-world outdoor noisy videos. Our code and dataset are available at https://github.com/cao-cong/RViDeformer.
Authors: Kun Huang, Linli Zhou, Shi Pu
Abstract: This paper proposes two distributed random reshuffling methods, namely Gradient Tracking with Random Reshuffling (GT-RR) and Exact Diffusion with Random Reshuffling (ED-RR), to solve the distributed optimization problem over a connected network, where a set of agents aim to minimize the average of their local cost functions. Both algorithms invoke random reshuffling (RR) update for each agent, inherit favorable characteristics of RR for minimizing smooth nonconvex objective functions, and improve the performance of previous distributed random reshuffling methods both theoretically and empirically. Specifically, both GT-RR and ED-RR achieve the convergence rate of $O(1/[(1-\lambda)^{1/3}m^{1/3}T^{2/3}])$ in driving the (minimum) expected squared norm of the gradient to zero, where $T$ denotes the number of epochs, $m$ is the sample size for each agent, and $1-\lambda$ represents the spectral gap of the mixing matrix. When the objective functions further satisfy the Polyak-{\L}ojasiewicz (PL) condition, we show GT-RR and ED-RR both achieve $O(1/[(1-\lambda)mT^2])$ convergence rate in terms of the averaged expected differences between the agents' function values and the global minimum value. Notably, both results are comparable to the convergence rates of centralized RR methods (up to constant factors depending on the network topology) and outperform those of previous distributed random reshuffling algorithms.
Authors: Nikhil Chandak, Shashwat Goel, Dominik Peters
Abstract: We study the problem of fair sequential decision making given voter preferences. In each round, a decision rule must choose a decision from a set of alternatives where each voter reports which of these alternatives they approve. Instead of going with the most popular choice in each round, we aim for proportional representation across rounds, using axioms inspired by the multi-winner voting literature. The axioms require that every group of $\alpha\%$ of the voters that agrees in every round (i.e., approves a common alternative), must approve at least $\alpha\%$ of the decisions. A stronger version of the axioms requires that every group of $\alpha\%$ of the voters that agrees in a $\beta$ fraction of rounds must approve $\beta\cdot\alpha\%$ of the decisions. We show that three attractive voting rules satisfy axioms of this style. One of them (Sequential Phragm\'en) makes its decisions online, and the other two satisfy strengthened versions of the axioms but make decisions semi-online (Method of Equal Shares) or fully offline (Proportional Approval Voting). We present empirical results for these rules based on synthetic data and U.S. political elections. We also run experiments using the moral machine dataset about ethical dilemmas: We train preference models on user responses from different countries and let the models cast votes. We find that aggregating these votes using our rules leads to a more equal utility distribution across demographics than making decisions using a single global preference model.
Authors: Robbert Reijnen, Igor G. Smit, Hongxiang Zhang, Yaoxin Wu, Zaharah Bukhsh, Yingqian Zhang
Abstract: Job shop scheduling problems address the routing and sequencing of tasks in a job shop setting. Despite significant interest from operations research and machine learning communities over the years, a comprehensive platform for testing and comparing solution methods has been notably lacking. To fill this gap, we introduce a unified implementation of job shop scheduling problems and their solution methods, addressing the long-standing need for a standardized benchmarking platform in this domain. Our platform supports classic Job Shop (JSP), Flow Shop (FSP), Flexible Job Shop (FJSP), and Assembly Job Shop (AJSP), as well as variants featuring Sequence-Dependent Setup Times (SDST), variants with online arrivals of jobs, and combinations of these problems (e.g., FJSP-SDST and FAJSP). The platfrom provides a wide range of scheduling solution methods, from heuristics, metaheuristics, and exact optimization to deep reinforcement learning. The implementation is available as an open-source GitHub repository, serving as a collaborative hub for researchers, practitioners, and those new to the field. Beyond enabling direct comparisons with existing methods on widely studied benchmark problems, this resource serves as a robust starting point for addressing constrained and complex problem variants. By establishing a comprehensive and unified foundation, this platform is designed to consolidate existing knowledge and to inspire the development of next-generation algorithms in job shop scheduling research.
Authors: Jin-Hong Du, Larry Wasserman, Kathryn Roeder
Abstract: Tens of thousands of simultaneous hypothesis tests are routinely performed in genomic studies to identify differentially expressed genes. However, due to unmeasured confounders, many standard statistical approaches may be substantially biased. This paper investigates the large-scale hypothesis testing problem for multivariate generalized linear models in the presence of confounding effects. Under arbitrary confounding mechanisms, we propose a unified statistical estimation and inference framework that harnesses orthogonal structures and integrates linear projections into three key stages. It begins by disentangling marginal and uncorrelated confounding effects to recover the latent coefficients. Subsequently, latent factors and primary effects are jointly estimated through lasso-type optimization. Finally, we incorporate projected and weighted bias-correction steps for hypothesis testing. Theoretically, we establish the identification conditions of various effects and non-asymptotic error bounds. We show effective Type-I error control of asymptotic $z$-tests as sample and response sizes approach infinity. Numerical experiments demonstrate that the proposed method controls the false discovery rate by the Benjamini-Hochberg procedure and is more powerful than alternative methods. By comparing single-cell RNA-seq counts from two groups of samples, we demonstrate the suitability of adjusting confounding effects when significant covariates are absent from the model.
Authors: Wanja de Sombre, Andrea Ortiz, Frank Aurzada, Anja Klein
Abstract: Popular methods to quantify transmitted data quality are the Age of Information (AoI), the Query Age of Information (QAoI), and the Age of Incorrect Information (AoII). We consider these metrics in a point-to-point wireless communication system, where the transmitter monitors a process and sends status updates to a receiver. The challenge is to decide on the best time for an update, balancing the transmission energy and the age-based metric at the receiver. Due to the inherent risk of high age-based metric values causing complications such as unstable system states, we introduce the new concept of risky states to denote states with high age-based metric. We use this new notion of risky states to quantify and minimize this risk of experiencing high age-based metrics by directly deriving the frequency of risky states as a novel risk-metric. Building on this foundation, we introduce two risk-sensitive strategies for AoI, QAoI and AoII. The first strategy uses system knowledge, i.e., channel quality and packet arrival probability, to find an optimal strategy that transmits when the age-based metric exceeds a tunable threshold. A lower threshold leads to higher risk-sensitivity. The second strategy uses an enhanced Q-learning approach and balances the age-based metric, the transmission energy and the frequency of risky states without requiring knowledge about the system. Numerical results affirm our risk-sensitive strategies' high effectiveness.
Authors: Shuheng Zhou
Abstract: In this paper, we consider the problem of partitioning a small data sample of size $n$ drawn from a mixture of 2 sub-gaussian distributions in $\mathbb{R}^p$. We consider semidefinite programming relaxations of an integer quadratic program that is formulated as finding the maximum cut on a graph, where edge weights in the cut represent dissimilarity scores between two nodes based on their $p$ features. We are interested in the case that individual features are of low average quality $\gamma$, and we want to use as few of them as possible to correctly partition the sample. Denote by $\Delta^2:=p \gamma$ the $\ell_2^2$ distance between two centers (mean vectors) in $\mathbb{R}^p$. The goal is to allow a full range of tradeoffs between $n, p, \gamma$ in the sense that partial recovery (success rate $< 100\%$) is feasible once the signal to noise ratio $s^2 := \min\{np \gamma^2, \Delta^2\}$ is lower bounded by a constant. For both balanced and unbalanced cases, we allow each population to have distinct covariance structures with diagonal matrices as special cases. In the present work, (a) we provide a unified framework for analyzing three computationally efficient algorithms, namely, SDP1, BalancedSDP, and Spectral clustering; and (b) we prove that the misclassification error decays exponentially with respect to the SNR $s^2$ for SDP1. Moreover, for balanced partitions, we design an estimator {\bf BalancedSDP} with a superb debiasing property. Indeed, with this new estimator, we remove an assumption (A2) on bounding the trace difference between the two population covariance matrices while proving the exponential error bound as stated above. These estimators and their statistical analyses are novel to the best of our knowledge. We provide simulation evidence illuminating the theoretical predictions.
Authors: Jinghuai Yao, Puyuan Du, Yucheng Zhao, Yubo Wang
Abstract: Visible (VIS) imagery is important for monitoring Tropical Cyclones (TCs) but is unavailable at night. This study presents a Conditional Generative Adversarial Networks (CGAN) model to generate nighttime VIS imagery with significantly enhanced accuracy and spatial resolution. Our method offers three key improvements compared to existing models. First, we replaced the L1 loss in the pix2pix framework with the Structural Similarity Index Measure (SSIM) loss, which significantly reduced image blurriness. Second, we selected multispectral infrared (IR) bands as input based on a thorough examination of their spectral properties, providing essential physical information for accurate simulation. Third, we incorporated the direction parameters of the sun and the satellite, which addressed the dependence of VIS images on sunlight directions and enabled a much larger training set from continuous daytime data. The model was trained and validated using data from the Advanced Himawari Imager (AHI) in the daytime, achieving statistical results of SSIM = 0.923 and Root Mean Square Error (RMSE) = 0.0299, which significantly surpasses existing models. We also performed a cross-satellite nighttime model validation using the Day/Night Band (DNB) of the Visible/Infrared Imager Radiometer Suite (VIIRS), which yields outstanding results compared to existing models. Our model is operationally applied to generate accurate VIS imagery with arbitrary virtual sunlight directions, significantly contributing to the nighttime monitoring of various meteorological phenomena.
Authors: Michael Gimelfarb, Ayal Taitler, Scott Sanner
Abstract: We propose the Constraint-Generation Policy Optimization (CGPO) framework to optimize policy parameters within compact and interpretable policy classes for mixed discrete-continuous Markov Decision Processes (DC-MDP). CGPO can not only provide bounded policy error guarantees over an infinite range of initial states for many DC-MDPs with expressive nonlinear dynamics, but it can also provably derive optimal policies in cases where it terminates with zero error. Furthermore, CGPO can generate worst-case state trajectories to diagnose policy deficiencies and provide counterfactual explanations of optimal actions. To achieve such results, CGPO proposes a bilevel mixed-integer nonlinear optimization framework for optimizing policies in defined expressivity classes (e.g. piecewise linear) and reduces it to an optimal constraint generation methodology that adversarially generates worst-case state trajectories. Furthermore, leveraging modern nonlinear optimizers, CGPO can obtain solutions with bounded optimality gap guarantees. We handle stochastic transitions through chance constraints, providing high-probability performance guarantees. We also present a roadmap for understanding the computational complexities of different expressivity classes of policy, reward, and transition dynamics. We experimentally demonstrate the applicability of CGPO across various domains, including inventory control, management of a water reservoir system, and physics control. In summary, CGPO provides structured, compact and explainable policies with bounded performance guarantees, enabling worst-case scenario generation and counterfactual policy diagnostics.
Authors: Yoshiaki Kitazawa
Abstract: Density ratio estimation (DRE) is a fundamental machine learning technique for capturing relationships between two probability distributions. State-of-the-art DRE methods estimate the density ratio using neural networks trained with loss functions derived from variational representations of $f$-divergences. However, existing methods face optimization challenges, such as overfitting due to lower-unbounded loss functions, biased mini-batch gradients, vanishing training loss gradients, and high sample requirements for Kullback--Leibler (KL) divergence loss functions. To address these issues, we focus on $\alpha$-divergence, which provides a suitable variational representation of $f$-divergence. Subsequently, a novel loss function for DRE, the $\alpha$-divergence loss function ($\alpha$-Div), is derived. $\alpha$-Div is concise but offers stable and effective optimization for DRE. The boundedness of $\alpha$-divergence provides the potential for successful DRE with data exhibiting high KL-divergence. Our numerical experiments demonstrate the effectiveness of $\alpha$-Div in optimization. However, the experiments also show that the proposed loss function offers no significant advantage over the KL-divergence loss function in terms of RMSE for DRE. This indicates that the accuracy of DRE is primarily determined by the amount of KL-divergence in the data and is less dependent on $\alpha$-divergence.
Authors: Xuandong Zhao, Lei Li, Yu-Xiang Wang
Abstract: In this paper, we propose a new decoding method called Permute-and-Flip (PF) decoder. It enjoys stability properties similar to the standard sampling decoder, but is provably up to 2x better in its quality-stability tradeoff than sampling and never worse than any other decoder. We also design a cryptographic watermarking scheme analogous to Aaronson (2023)'s Gumbel watermark, but naturally tailored for PF decoder. The watermarking scheme does not change the distribution to sample, while allowing arbitrarily low false positive rate and high recall whenever the generated text has high entropy. Our experiments show that the PF decoder (and its watermarked counterpart) significantly outperform(s) naive sampling (and its Gumbel watermarked counterpart) in terms of perplexity, while retaining the same stability (and detectability), hence making it a promising new approach for LLM decoding. The code is available at https://github.com/XuandongZhao/pf-decoding
Authors: Feiyang Wu, Xavier Nal, Jaehwi Jang, Wei Zhu, Zhaoyuan Gu, Anqi Wu, Ye Zhao
Abstract: Humanoid robots promise transformative capabilities for industrial and service applications. While recent advances in Reinforcement Learning (RL) yield impressive results in locomotion, manipulation, and navigation, the proposed methods typically require enormous simulation samples to account for real-world variability. This work proposes a novel one-stage training framework-Learn to Teach (L2T)-which unifies teacher and student policy learning. Our approach recycles simulator samples and synchronizes the learning trajectories through shared dynamics, significantly reducing sample complexities and training time while achieving state-of-the-art performance. Furthermore, we validate the RL variant (L2T-RL) through extensive simulations and hardware tests on the Digit robot, demonstrating zero-shot sim-to-real transfer and robust performance over 12+ challenging terrains without depth estimation modules.
Authors: Elad Aigner-Horev, Daniel Rosenberg, Roi Weiss
Abstract: The resilience of a Rademacher chaos is the maximum number of adversarial sign-flips that the chaos can sustain without having its largest atom probability significantly altered. Inspired by probabilistic lower-bound guarantees for the resilience of linear Rademacher chaos, obtained by Bandeira, Ferber, and Kwan (Advances in Mathematics, Vol. $319$, $2017$), we provide probabilistic lower-bound guarantees for the resilience of Rademacher chaos of arbitrary yet sufficiently low degree. Our main results distinguish between Rademacher chaos of order two and those of higher order. In that, our first main result pertains to the resilience of decoupled bilinear Rademacher forms where different asymptotic behaviour is observed for sparse and dense matrices. For our second main result, we bootstrap our first result in order to provide resilience guarantees for quadratic Rademacher chaos. Our third main result, generalises the first and handles the resilience of decoupled Rademacher chaos of arbitrary yet sufficiently low order. Our results for decoupled Rademacher chaos of order two and that of higher order whilst are established through the same conceptual framework, differ substantially. A difference incurred due to the implementation of the same conceptual argument. The order two result is established using Dudley's maximal inequality for sub-Gaussian processes, the Hanson-Wright inequality, as well as the Kolmogorov-Rogozin inequality. To handle higher order chaos, appeals to Dudley's inequality as well as the Hanson-Wright inequality are replaced with tools suited for random tensors. Appeals to the Hanson-Wright inequality are replaced with appeals to a concentration result for random tensors put forth by Adamczak and Wolff. Our results are instance-dependent and thus allow for the efficient computation of resilience guarantees provided the order of the chaos is constant.
Authors: Jianming Lv, Chengjun Wang, Depin Liang, Qianli Ma, Wei Chen, Xueqi Cheng
Abstract: Utilizing unlabeled data in the target domain to perform continuous optimization is critical to enhance the generalization ability of neural networks. Most domain adaptation methods focus on time-consuming optimization of deep feature extractors, which limits the deployment on lightweight edge devices. Inspired by the memory mechanism and powerful generalization ability of biological neural networks in human brains, we propose a novel gradient-free Elastic Memory Network, namely EMN, to support quick fine-tuning of the mapping between features and prediction without heavy optimization of deep features. In particular, EMN adopts randomly connected neurons to memorize the association of features and labels, where the signals in the network are propagated as impulses, and the prediction is made by associating the memories stored on neurons based on their confidence. More importantly, EMN supports reinforced memorization of feature mapping based on unlabeled data to quickly adapt to a new domain. Experiments based on four cross-domain real-world datasets show that EMN can achieve up to 10% enhancement of performance while only needing less than 1% timing cost of traditional domain adaptation methods.
Authors: Xin Wang, Yu Zheng, Zhongwei Wan, Mi Zhang
Abstract: The advancements in Large Language Models (LLMs) have been hindered by their substantial sizes, which necessitates LLM compression methods for practical deployment. Singular Value Decomposition (SVD) offers a promising solution for LLM compression. However, state-of-the-art SVD-based LLM compression methods have two key limitations: truncating smaller singular values may lead to higher compression loss, and the lack of update on the compressed weights after SVD truncation. In this work, we propose SVD-LLM, a SVD-based post-training LLM compression method that addresses the limitations of existing methods. SVD-LLM incorporates a truncation-aware data whitening technique to ensure a direct mapping between singular values and compression loss. Moreover, SVD-LLM adopts a parameter update with sequential low-rank approximation to compensate for the accuracy degradation after SVD compression. We evaluate SVD-LLM on 10 datasets and seven models from three different LLM families at three different scales. Our results demonstrate the superiority of SVD-LLM over state-of-the-arts, especially at high model compression ratios. Our code is available at https://github.com/AIoT-MLSys-Lab/SVD-LLM
Authors: Yang Cai, Constantinos Daskalakis, Haipeng Luo, Chen-Yu Wei, Weiqiang Zheng
Abstract: While Online Gradient Descent and other no-regret learning procedures are known to efficiently converge to a coarse correlated equilibrium in games where each agent's utility is concave in their own strategy, this is not the case when utilities are non-concave -- a common scenario in machine learning applications involving strategies parameterized by deep neural networks, or when agents' utilities are computed by neural networks, or both. Non-concave games introduce significant game-theoretic and optimization challenges: (i) Nash equilibria may not exist; (ii) local Nash equilibria, though they exist, are intractable; and (iii) mixed Nash, correlated, and coarse correlated equilibria generally have infinite support and are intractable. To sidestep these challenges, we revisit the classical solution concept of $\Phi$-equilibria introduced by Greenwald and Jafari [2003], which is guaranteed to exist for an arbitrary set of strategy modifications $\Phi$ even in non-concave games [Stolz and Lugosi, 2007]. However, the tractability of $\Phi$-equilibria in such games remains elusive. In this paper, we initiate the study of tractable $\Phi$-equilibria in non-concave games and examine several natural families of strategy modifications. We show that when $\Phi$ is finite, there exists an efficient uncoupled learning algorithm that converges to the corresponding $\Phi$-equilibria. Additionally, we explore cases where $\Phi$ is infinite but consists of local modifications. We show that approximating local $\Phi$-equilibria beyond the first-order stationary regime is computationally intractable. In contrast, within this regime, we show Online Gradient Descent efficiently converges to $\Phi$-equilibria for several natural infinite families of modifications, including a new structural family of modifications inspired by the well-studied proximal operator.
Authors: Sebastian Doerrich, Francesco Di Salvo, Julius Brockmann, Christian Ledig
Abstract: The integration of deep learning based systems in clinical practice is often impeded by challenges rooted in limited and heterogeneous medical datasets. In addition, the field has increasingly prioritized marginal performance gains on a few, narrowly scoped benchmarks over clinical applicability, slowing down meaningful algorithmic progress. This trend often results in excessive fine-tuning of existing methods on selected datasets rather than fostering clinically relevant innovations. In response, this work introduces a comprehensive benchmark for the MedMNIST+ dataset collection, designed to diversify the evaluation landscape across several imaging modalities, anatomical regions, classification tasks and sample sizes. We systematically reassess commonly used Convolutional Neural Networks (CNNs) and Vision Transformer (ViT) architectures across distinct medical datasets, training methodologies, and input resolutions to validate and refine existing assumptions about model effectiveness and development. Our findings suggest that computationally efficient training schemes and modern foundation models offer viable alternatives to costly end-to-end training. Additionally, we observe that higher image resolutions do not consistently improve performance beyond a certain threshold. This highlights the potential benefits of using lower resolutions, particularly in prototyping stages, to reduce computational demands without sacrificing accuracy. Notably, our analysis reaffirms the competitiveness of CNNs compared to ViTs, emphasizing the importance of comprehending the intrinsic capabilities of different architectures. Finally, by establishing a standardized evaluation framework, we aim to enhance transparency, reproducibility, and comparability within the MedMNIST+ dataset collection. Code is available at https://github.com/sdoerrich97/rethinking-model-prototyping-MedMNISTPlus .
URLs: https://github.com/sdoerrich97/rethinking-model-prototyping-MedMNISTPlus
Authors: Kabir Ahuja, Vidhisha Balachandran, Madhur Panwar, Tianxing He, Noah A. Smith, Navin Goyal, Yulia Tsvetkov
Abstract: Transformers trained on natural language data have been shown to learn its hierarchical structure and generalize to sentences with unseen syntactic structures without explicitly encoding any structural bias. In this work, we investigate sources of inductive bias in transformer models and their training that could cause such generalization behavior to emerge. We extensively experiment with transformer models trained on multiple synthetic datasets and with different training objectives and show that while other objectives e.g. sequence-to-sequence modeling, prefix language modeling, often failed to lead to hierarchical generalization, models trained with the language modeling objective consistently learned to generalize hierarchically. We then conduct pruning experiments to study how transformers trained with the language modeling objective encode hierarchical structure. When pruned, we find joint existence of subnetworks within the model with different generalization behaviors (subnetworks corresponding to hierarchical structure and linear order). Finally, we take a Bayesian perspective to further uncover transformers' preference for hierarchical generalization: We establish a correlation between whether transformers generalize hierarchically on a dataset and whether the simplest explanation of that dataset is provided by a hierarchical grammar compared to regular grammars exhibiting linear generalization.
Authors: Kehan Shi, Martin Burger
Abstract: As a generalization of graphs, hypergraphs are widely used to model higher-order relations in data. This paper explores the benefit of the hypergraph structure for the interpolation of point cloud data that contain no explicit structural information. We define the $\varepsilon_n$-ball hypergraph and the $k_n$-nearest neighbor hypergraph on a point cloud and study the $p$-Laplacian regularization on the hypergraphs. We prove the variational consistency between the hypergraph $p$-Laplacian regularization and the continuum $p$-Laplacian regularization in a semisupervised setting when the number of points $n$ goes to infinity while the number of labeled points remains fixed. A key improvement compared to the graph case is that the results rely on weaker assumptions on the upper bound of $\varepsilon_n$ and $k_n$. To solve the convex but non-differentiable large-scale optimization problem, we utilize the stochastic primal-dual hybrid gradient algorithm. Numerical experiments on data interpolation verify that the hypergraph $p$-Laplacian regularization outperforms the graph $p$-Laplacian regularization in preventing the development of spikes at the labeled points.
Authors: Ali Bahri, Moslem Yazdanpanah, Mehrdad Noori, Milad Cheraghalikhani, Gustavo Adolfo Vargas Hakim, David Osowiechi, Farzad Beizaee, Ismail Ben Ayed, Christian Desrosiers
Abstract: We introduce a pioneering approach to self-supervised learning for point clouds, employing a geometrically informed mask selection strategy called GeoMask3D (GM3D) to boost the efficiency of Masked Auto Encoders (MAE). Unlike the conventional method of random masking, our technique utilizes a teacher-student model to focus on intricate areas within the data, guiding the model's focus toward regions with higher geometric complexity. This strategy is grounded in the hypothesis that concentrating on harder patches yields a more robust feature representation, as evidenced by the improved performance on downstream tasks. Our method also presents a complete-to-partial feature-level knowledge distillation technique designed to guide the prediction of geometric complexity utilizing a comprehensive context from feature-level information. Extensive experiments confirm our method's superiority over State-Of-The-Art (SOTA) baselines, demonstrating marked improvements in classification, and few-shot tasks.
Authors: Nadav Timor, Jonathan Mamou, Daniel Korat, Moshe Berchansky, Oren Pereg, Moshe Wasserblat, Tomer Galanti, Michal Gordon, David Harel
Abstract: This paper introduces distributed speculative inference (DSI), a novel inference algorithm that is provably faster than speculative inference (SI) [leviathan2023, chen2023, miao2024, sun2025, timor2025] and standard autoregressive inference (non-SI). Like other SI algorithms, DSI operates on frozen language models (LMs), requiring no training or architectural modifications, and it preserves the target distribution. Prior studies on SI have demonstrated empirical speedups over non-SI--but rely on sufficiently fast and accurate drafters, which are often unavailable in practice. We identify a gap where SI can be slower than non-SI if drafters are too slow or inaccurate. We close this gap by proving that DSI is faster than both SI and non-SI--given any drafters. DSI is therefore not only faster than SI, but also unlocks the acceleration of LMs for which SI fails. DSI leverages speculation parallelism (SP), a novel type of task parallelism, to orchestrate target and drafter instances that overlap in time, establishing a new foundational tradeoff between computational resources and latency. Our simulations show that DSI is 1.29-1.92x faster than SI in single-node setups for various off-the-shelf LMs and tasks. We open-source all our code.
Authors: Haya Diwan, Jinrui Gou, Cameron Musco, Christopher Musco, Torsten Suel
Abstract: There has been significant recent interest in graph-based nearest neighbor search methods, many of which are centered on the construction of navigable graphs over high-dimensional point sets. A graph is navigable if we can successfully move from any starting node to any target node using a greedy routing strategy where we always move to the neighbor that is closest to the destination according to a given distance function. The complete graph is navigable for any point set, but the important question for applications is if sparser graphs can be constructed. While this question is fairly well understood in low-dimensions, we establish some of the first upper and lower bounds for high-dimensional point sets. First, we give a simple and efficient way to construct a navigable graph with average degree $O(\sqrt{n \log n })$ for any set of $n$ points, in any dimension, for any distance function. We compliment this result with a nearly matching lower bound: even under the Euclidean metric in $O(\log n)$ dimensions, a random point set has no navigable graph with average degree $O(n^{\alpha})$ for any $\alpha < 1/2$. Our lower bound relies on sharp anti-concentration bounds for binomial random variables, which we use to show that the near-neighborhoods of a set of random points do not overlap significantly, forcing any navigable graph to have many edges.
Authors: Yulong Yang, Felix O'Mahony, Christine Allen-Blanchette
Abstract: In this paper, we introduce group convolutional neural networks (GCNNs) equivariant to color variation. GCNNs have been designed for a variety of geometric transformations from 2D and 3D rotation groups, to semi-groups such as scale. Despite the improved interpretability, accuracy and generalizability of these architectures, GCNNs have seen limited application in the context of perceptual quantities. Notably, the recent CEConv network uses a GCNN to achieve equivariance to hue transformations by convolving input images with a hue rotated RGB filter. However, this approach leads to invalid RGB values which break equivariance and degrade performance. We resolve these issues with a lifting layer that transforms the input image directly, thereby circumventing the issue of invalid RGB values and improving equivariance error by over three orders of magnitude. Moreover, we extend the notion of color equivariance to include equivariance to saturation and luminance shift. Our hue-, saturation-, luminance- and color-equivariant networks achieve strong generalization to out-of-distribution perceptual variations and improved sample efficiency over conventional architectures. We demonstrate the utility of our approach on synthetic and real world datasets where we consistently outperform competitive baselines.
Authors: Purva Pruthi, David Jensen
Abstract: Many real-world systems can be usefully represented as sets of interacting components. Examples include computational systems, such as query processors and compilers, natural systems, such as cells and ecosystems, and social systems, such as families and organizations. However, current approaches to estimating potential outcomes and causal effects typically treat such systems as single units, represent them with a fixed set of variables, and assume a homogeneous data-generating process. In this work, we study a compositional approach for estimating individual-level potential outcomes and causal effects in structured systems, where each unit is represented by an instance-specific composition of multiple heterogeneous components. The compositional approach decomposes unit-level causal queries into more fine-grained queries, explicitly modeling how unit-level interventions affect component-level outcomes to generate a unit's outcome. We demonstrate this approach using modular neural network architectures and show that it provides benefits for causal effect estimation from observational data, such as accurate causal effect estimation for structured units, increased sample efficiency, improved overlap between treatment and control groups, and compositional generalization to units with unseen combinations of components. Remarkably, our results show that compositional modeling can improve the accuracy of causal estimation even when component-level outcomes are unobserved. We also create and use a set of real-world evaluation environments for the empirical evaluation of compositional approaches for causal effect estimation and demonstrate the role of composition structure, varying amounts of component-level data access, and component heterogeneity in the performance of compositional models as compared to the non-compositional approaches.
Authors: Lijie Hu, Tianhao Huang, Huanyi Xie, Chenyang Ren, Zhengyu Hu, Lu Yu, Di Wang
Abstract: Concept Bottleneck Models (CBMs) have garnered increasing attention due to their ability to provide concept-based explanations for black-box deep learning models while achieving high final prediction accuracy using human-like concepts. However, the training of current CBMs heavily relies on the accuracy and richness of annotated concepts in the dataset. These concept labels are typically provided by experts, which can be costly and require significant resources and effort. Additionally, concept saliency maps frequently misalign with input saliency maps, causing concept predictions to correspond to irrelevant input features - an issue related to annotation alignment. To address these limitations, we propose a new framework called SSCBM (Semi-supervised Concept Bottleneck Model). Our SSCBM is suitable for practical situations where annotated data is scarce. By leveraging joint training on both labeled and unlabeled data and aligning the unlabeled data at the concept level, we effectively solve these issues. We proposed a strategy to generate pseudo labels and an alignment loss. Experiments demonstrate that our SSCBM is both effective and efficient. With only 20% labeled data, we achieved 93.19% (96.39% in a fully supervised setting) concept accuracy and 75.51% (79.82% in a fully supervised setting) prediction accuracy.
Authors: Ioannis Tsiamas, Santiago Pascual, Chunghsin Yeh, Joan Serr\`a
Abstract: Contrastive learning has emerged as a powerful technique in audio-visual representation learning, leveraging the natural co-occurrence of audio and visual modalities in webscale video datasets. However, conventional contrastive audio-visual learning (CAV) methodologies often rely on aggregated representations derived through temporal aggregation, neglecting the intrinsic sequential nature of the data. This oversight raises concerns regarding the ability of standard approaches to capture and utilize fine-grained information within sequences. In response to this limitation, we propose sequential contrastive audiovisual learning (SCAV), which contrasts examples based on their non-aggregated representation space using multidimensional sequential distances. Audio-visual retrieval experiments with the VGGSound and Music datasets demonstrate the effectiveness of SCAV, with up to 3.5x relative improvements in recall against traditional aggregation-based contrastive learning and other previously proposed methods, which utilize more parameters and data. We also show that models trained with SCAV exhibit a significant degree of flexibility regarding the metric employed for retrieval, allowing us to use a hybrid retrieval approach that is both effective and efficient.
Authors: Chihcheng Hsieh, Catarina Moreira, Isabel Blanco Nobre, Sandra Costa Sousa, Chun Ouyang, Margot Brereton, Joaquim Jorge, Jacinto C. Nascimento
Abstract: X-ray images are vital in medical diagnostics, but their effectiveness is limited without clinical context. Radiologists often find chest X-rays insufficient for diagnosing underlying diseases, necessitating the integration of structured clinical features with radiology reports. To address this, we introduce DALL-M, a novel framework that enhances clinical datasets by generating contextual synthetic data. DALL-M augments structured patient data, including vital signs (e.g., heart rate, oxygen saturation), radiology findings (e.g., lesion presence), and demographic factors. It integrates this tabular data with contextual knowledge extracted from radiology reports and domain-specific resources (e.g., Radiopaedia, Wikipedia), ensuring clinical consistency and reliability. DALL-M follows a three-phase process: (i) clinical context storage, (ii) expert query generation, and (iii) context-aware feature augmentation. Using large language models (LLMs), it generates both contextual synthetic values for existing clinical features and entirely new, clinically relevant features. Applied to 799 cases from the MIMIC-IV dataset, DALL-M expanded the original 9 clinical features to 91. Empirical validation with machine learning models (including Decision Trees, Random Forests, XGBoost, and TabNET) demonstrated a 16.5% improvement in F1 score and a 25% increase in Precision and Recall. DALL-M bridges an important gap in clinical data augmentation by preserving data integrity while enhancing predictive modeling in healthcare. Our results show that integrating LLM-generated synthetic features significantly improves model performance, making DALL-M a scalable and practical approach for AI-driven medical diagnostics.
Authors: Tural Mammadov, Dietrich Klakow, Alexander Koller, Andreas Zeller
Abstract: We introduce Modelizer - a novel framework that, given a black-box program, learns a model from its input/output behavior using neural machine translation algorithms. The resulting model mocks the original program: Given an input, the model predicts the output that would have been produced by the program. However, the model is also reversible - that is, the model can predict the input that would have produced a given output. Finally, the model is differentiable and can be efficiently restricted to predict only a certain aspect of the program behavior. Modelizer uses grammars to synthesize and inputs and unsupervised tokenizers to decompose the resulting outputs, allowing it to learn sequence-to-sequence associations between token streams. Other than input grammars, Modelizer only requires the ability to execute the program. The resulting models are small, requiring fewer than 6.3 million parameters for languages such as Markdown or HTML; and they are accurate, achieving up to 95.4% accuracy and a BLEU score of 0.98 with standard error 0.04 in mocking real-world applications. As it learns from and predicts executions rather than code, Modelizer departs from the LLM-centric research trend, opening new opportunities for program-specific models that are fully tuned towards individual programs. Indeed, we foresee several applications of these models, especially as the output of the program can be any aspect of program behavior. Beyond mocking and predicting program behavior, the models can also synthesize inputs that are likely to produce a particular behavior, such as failures or coverage, thus assisting in program understanding and maintenance.
Authors: Zhichao Han, Mohit Pundir, Olga Fink, David S. Kammer
Abstract: Accurately modeling the mechanical behavior of materials is crucial for numerous engineering applications. The quality of these models depends directly on the accuracy of the constitutive law that defines the stress-strain relation. However, discovering these constitutive material laws remains a significant challenge, in particular when only material deformation data is available. To address this challenge, unsupervised machine learning methods have been proposed to learn the constitutive law from deformation data. Nonetheless, existing approaches have several limitations: they either fail to ensure that the learned constitutive relations are consistent with physical principles, or they rely on boundary force data for training which are unavailable in many in-situ scenarios. Here, we introduce a machine learning approach to learn physics-consistent constitutive relations solely from material deformation without boundary force information. This is achieved by considering a dynamic formulation rather than static equilibrium data and applying an input convex neural network (ICNN). We validate the effectiveness of the proposed method on a diverse range of hyperelastic material laws. We demonstrate that it is robust to a significant level of noise and that it converges to the ground truth with increasing data resolution. We also show that the model can be effectively trained using a displacement field from a subdomain of the test specimen and that the learned constitutive relation from one material sample is transferable to other samples with different geometries. The developed methodology provides an effective tool for discovering constitutive relations. It is, due to its design based on dynamics, particularly suited for applications to strain-rate-dependent materials and situations where constitutive laws need to be inferred from in-situ measurements without access to global force data.
Authors: Danfeng Guo, Demetri Terzopoulos
Abstract: Large Vision-Language Models (LVLMs) have achieved significant success in recent years, and they have been extended to the medical domain. Although demonstrating satisfactory performance on medical Visual Question Answering (VQA) tasks, Medical LVLMs (MLVLMs) suffer from the hallucination problem, which makes them fail to diagnose complex pathologies. Moreover, they readily fail to learn minority pathologies due to imbalanced training data. We propose two prompting strategies for MLVLMs that reduce hallucination and improve VQA performance. In the first strategy, we provide a detailed explanation of the queried pathology. In the second strategy, we fine-tune a cheap, weak learner to achieve high performance on a specific metric, and textually provide its judgment to the MLVLM. Tested on the MIMIC-CXR-JPG and Chexpert datasets, our methods significantly improve the diagnostic F1 score, with the highest increase being 0.27. We also demonstrate that our prompting strategies can be extended to general LVLM domains. Based on POPE metrics, it effectively suppresses the false negative predictions of existing LVLMs and improves Recall by approximately 0.07.
Authors: Moussa Koulako Bala Doumbouya, Ananjan Nandi, Gabriel Poesia, Davide Ghilardi, Anna Goldie, Federico Bianchi, Dan Jurafsky, Christopher D. Manning
Abstract: Despite their demonstrated valuable capabilities, state-of-the-art (SOTA) widely deployed large language models (LLMs) still have the potential to cause harm to society due to the ineffectiveness of their safety filters, which can be bypassed by prompt transformations called jailbreak attacks. Current approaches to LLM safety assessment, which employ datasets of templated prompts and benchmarking pipelines, fail to cover sufficiently large and diverse sets of jailbreak attacks, leading to the widespread deployment of unsafe LLMs. Recent research showed that novel jailbreak attacks could be derived by composition; however, a formal composable representation for jailbreak attacks, which, among other benefits, could enable the exploration of a large compositional space of jailbreak attacks through program synthesis methods, has not been previously proposed. We introduce h4rm3l, a novel approach that addresses this gap with a human-readable domain-specific language (DSL). Our framework comprises: (1) The h4rm3l DSL, which formally expresses jailbreak attacks as compositions of parameterized string transformation primitives. (2) A synthesizer with bandit algorithms that efficiently generates jailbreak attacks optimized for a target black box LLM. (3) The h4rm3l red-teaming software toolkit that employs the previous two components and an automated harmful LLM behavior classifier that is strongly aligned with human judgment. We demonstrate h4rm3l's efficacy by synthesizing a dataset of 2656 successful novel jailbreak attacks targeting 6 SOTA open-source and proprietary LLMs, and by benchmarking those models against a subset of these synthesized attacks. Our results show that h4rm3l's synthesized attacks are diverse and more successful than existing jailbreak attacks in literature, with success rates exceeding 90% on SOTA LLMs.
Authors: Caixing Wang, Xingdong Feng
Abstract: The random feature (RF) approach is a well-established and efficient tool for scalable kernel methods, but existing literature has primarily focused on kernel ridge regression with random features (KRR-RF), which has limitations in handling heterogeneous data with heavy-tailed noises. This paper presents a generalization study of kernel quantile regression with random features (KQR-RF), which accounts for the non-smoothness of the check loss in KQR-RF by introducing a refined error decomposition and establishing a novel connection between KQR-RF and KRR-RF. Our study establishes the capacity-dependent learning rates for KQR-RF under mild conditions on the number of RFs, which are minimax optimal up to some logarithmic factors. Importantly, our theoretical results, utilizing a data-dependent sampling strategy, can be extended to cover the agnostic setting where the target quantile function may not precisely align with the assumed kernel space. By slightly modifying our assumptions, the capacity-dependent error analysis can also be applied to cases with Lipschitz continuous losses, enabling broader applications in the machine learning community. To validate our theoretical findings, simulated experiments and a real data application are conducted.
Authors: Phillip Long, Zachary Novack, Taylor Berg-Kirkpatrick, Julian McAuley
Abstract: The recent explosion of generative AI-Music systems has raised numerous concerns over data copyright, licensing music from musicians, and the conflict between open-source AI and large prestige companies. Such issues highlight the need for publicly available, copyright-free musical data, in which there is a large shortage, particularly for symbolic music data. To alleviate this issue, we present PDMX: a large-scale open-source dataset of over 250K public domain MusicXML scores collected from the score-sharing forum MuseScore, making it the largest available copyright-free symbolic music dataset to our knowledge. PDMX additionally includes a wealth of both tag and user interaction metadata, allowing us to efficiently analyze the dataset and filter for high quality user-generated scores. Given the additional metadata afforded by our data collection process, we conduct multitrack music generation experiments evaluating how different representative subsets of PDMX lead to different behaviors in downstream models, and how user-rating statistics can be used as an effective measure of data quality. Examples can be found at https://pnlong.github.io/PDMX.demo/.
Authors: Sara Avesani, R\"udiger Kempf, Michael Multerer, Holger Wendland
Abstract: We study multiscale scattered data interpolation schemes for globally supported radial basis functions with focus on the Mat\'ern class. The multiscale approximation is constructed through a sequence of residual corrections, where radial basis functions with different lengthscale parameters are combined to capture varying levels of detail. We prove that the condition numbers of the the diagonal blocks of the corresponding multiscale system remain bounded independently of the particular level, allowing us to use an iterative solver with a bounded number of iterations for the numerical solution. Employing an appropriate diagonal scaling, the multiscale system becomes well conditioned. We exploit this fact to derive a general error estimate bounding the consistency error issuing from a numerical approximation of the multiscale system. To apply the multiscale approach to large data sets, we suggest to represent each level of the multiscale system in samplet coordinates. Samplets are localized, discrete signed measures exhibiting vanishing moments and allow for the sparse approximation of generalized Vandermonde matrices issuing from a vast class of radial basis functions. Given a quasi-uniform set of $N$ data sites, and local approximation spaces with exponentially decreasing dimension, the samplet compressed multiscale system can be assembled with cost $\mathcal{O}(N \log^2 N)$. The overall cost of the proposed approach is $\mathcal{O}(N \log^2 N)$. The theoretical findings are accompanied by extensive numerical studies in two and three spatial dimensions.
Authors: Jun Liu, Geng Yuan, Weihao Zeng, Hao Tang, Wenbin Zhang, Xue Lin, XiaoLin Xu, Dong Huang, Yanzhi Wang
Abstract: In research findings, co-deletion of the 1p/19q gene is associated with clinical outcomes in low-grade gliomas. The ability to predict 1p19q status is critical for treatment planning and patient follow-up. This study aims to utilize a specially MRI-based convolutional neural network for brain cancer detection. Although public networks such as RestNet and AlexNet can effectively diagnose brain cancers using transfer learning, the model includes quite a few weights that have nothing to do with medical images. As a result, the diagnostic results are unreliable by the transfer learning model. To deal with the problem of trustworthiness, we create the model from the ground up, rather than depending on a pre-trained model. To enable flexibility, we combined convolution stacking with a dropout and full connect operation, it improved performance by reducing overfitting. During model training, we also supplement the given dataset and inject Gaussian noise. We use three--fold cross-validation to train the best selection model. Comparing InceptionV3, VGG16, and MobileNetV2 fine-tuned with pre-trained models, our model produces better results. On an validation set of 125 codeletion vs. 31 not codeletion images, the proposed network achieves 96.37\% percent F1-score, 97.46\% percent precision, and 96.34\% percent recall when classifying 1p/19q codeletion and not codeletion images.
Authors: Yu Ying Chiu, Liwei Jiang, Yejin Choi
Abstract: As users increasingly seek guidance from LLMs for decision-making in daily life, many of these decisions are not clear-cut and depend significantly on the personal values and ethical standards of people. We present DailyDilemmas, a dataset of 1,360 moral dilemmas encountered in everyday life. Each dilemma presents two possible actions, along with affected parties and relevant human values for each action. Based on these dilemmas, we gather a repository of human values covering diverse everyday topics, such as interpersonal relationships, workplace, and environmental issues. With DailyDilemmas, we evaluate LLMs on these dilemmas to determine what action they will choose and the values represented by these action choices. Then, we analyze values through the lens of five theoretical frameworks inspired by sociology, psychology, and philosophy, including the World Values Survey, Moral Foundations Theory, Maslow's Hierarchy of Needs, Aristotle's Virtues, and Plutchik's Wheel of Emotions. For instance, we find LLMs are most aligned with self-expression over survival in World Values Survey and care over loyalty in Moral Foundations Theory. Interestingly, we find substantial preference differences in models for some core values. For example, for truthfulness, Mixtral-8x7B neglects it by 9.7% while GPT-4-turbo selects it by 9.4%. We also study the recent guidance released by OpenAI (ModelSpec), and Anthropic (Constitutional AI) to understand how their designated principles reflect their models' actual value prioritization when facing nuanced moral reasoning in daily-life settings. Finally, we find that end users cannot effectively steer such prioritization using system prompts.
Authors: Alihan H\"uy\"uk, Xinnuo Xu, Jacqueline Maasch, Aditya V. Nori, Javier Gonz\'alez
Abstract: Despite the increasing effectiveness of language models, their reasoning capabilities remain underdeveloped. In particular, causal reasoning through counterfactual question answering is lacking. This work aims to bridge this gap. We first derive novel metrics that balance accuracy in factual and counterfactual questions, capturing a more complete view of the reasoning abilities of language models than traditional factual-only based metrics. Second, we propose several fine-tuning approaches that aim to elicit better reasoning mechanisms, in the sense of the proposed metrics. Finally, we evaluate the performance of the fine-tuned language models in a variety of realistic scenarios. In particular, we investigate to what extent our fine-tuning approaches systemically achieve better generalization with respect to the base models in several problems that require, among others, inductive and deductive reasoning capabilities.
Authors: Yang Jin, Zhicheng Sun, Ningyuan Li, Kun Xu, Kun Xu, Hao Jiang, Nan Zhuang, Quzhe Huang, Yang Song, Yadong Mu, Zhouchen Lin
Abstract: Video generation requires modeling a vast spatiotemporal space, which demands significant computational resources and data usage. To reduce the complexity, the prevailing approaches employ a cascaded architecture to avoid direct training with full resolution latent. Despite reducing computational demands, the separate optimization of each sub-stage hinders knowledge sharing and sacrifices flexibility. This work introduces a unified pyramidal flow matching algorithm. It reinterprets the original denoising trajectory as a series of pyramid stages, where only the final stage operates at the full resolution, thereby enabling more efficient video generative modeling. Through our sophisticated design, the flows of different pyramid stages can be interlinked to maintain continuity. Moreover, we craft autoregressive video generation with a temporal pyramid to compress the full-resolution history. The entire framework can be optimized in an end-to-end manner and with a single unified Diffusion Transformer (DiT). Extensive experiments demonstrate that our method supports generating high-quality 5-second (up to 10-second) videos at 768p resolution and 24 FPS within 20.7k A100 GPU training hours. All code and models are open-sourced at https://pyramid-flow.github.io.
Authors: Hengxiang Zhang, Songxin Zhang, Bingyi Jing, Hongxin Wei
Abstract: In the era of large language models (LLMs), detecting pretraining data has been increasingly important due to concerns about fair evaluation and ethical risks. Current methods differentiate members and non-members by designing scoring functions, like Perplexity and Min-k%. However, the diversity and complexity of training data magnifies the difficulty of distinguishing, leading to suboptimal performance in detecting pretraining data. In this paper, we first explore the benefits of unseen data, which can be easily collected after the release of the LLM. We find that the perplexities of LLMs shift differently for members and non-members, after fine-tuning with a small amount of previously unseen data. In light of this, we introduce a novel and effective method termed Fine-tuned Score Deviation(FSD), which improves the performance of current scoring functions for pretraining data detection. In particular, we propose to measure the deviation distance of current scores after fine-tuning on a small amount of unseen data within the same domain. In effect, using a few unseen data can largely decrease the scores of all non-members, leading to a larger deviation distance than members. Extensive experiments demonstrate the effectiveness of our method, significantly improving the AUC score on common benchmark datasets across various models.
Authors: Zixin Wang, Dong Gong, Sen Wang, Zi Huang, Yadan Luo
Abstract: Contrastive Language-Image Pretraining (CLIP) excels at learning generalizable image representations but often falls short in zero-shot inference on certain downstream datasets. Test-time adaptation (TTA) mitigates this issue by adjusting components like normalization layers or context prompts, yet it typically requires large batch sizes and extensive augmentations, leading to high computational costs. This raises a key question: Can VLMs' performance drop in specific test cases be mitigated through efficient, training-free approaches? To explore the solution, we investigate token condensation (TC) techniques, originally designed to enhance vision transformer efficiency by refining token usage during inference. We observe that informative tokens improve visual-text alignment in VLMs like CLIP on unseen datasets. However, existing TC methods often fail to maintain in-distribution performance when reducing tokens, prompting us to ask: How can we transform TC into an effective ``free-lunch'' adaptation strategy for VLMs? To address this, we propose Token Condensation as Adaptation (TCA), a training-free adaptation method that takes a step beyond standard TC. Rather than passively discarding tokens, TCA condenses token representation by introducing reservoir-based domain anchor tokens for information-preserving token reduction and logits correction. TCA achieves up to a 21.4% performance improvement over the strongest baseline on cross-dataset benchmark and the CIFAR-100-Corrupted dataset while reducing GFLOPs by 12.2% to 48.9%, with minimal hyperparameter dependency on both CLIP and SigLIP series.
Authors: Talal Alrawajfeh, Joonas J\"alk\"o, Antti Honkela
Abstract: Differential privacy (DP) provides robust privacy guarantees for statistical inference, but this can lead to unreliable results and biases in downstream applications. While several noise-aware approaches have been proposed which integrate DP perturbation into the inference, they are limited to specific types of simple probabilistic models. In this work, we propose a novel method for noise-aware approximate Bayesian inference based on stochastic gradient variational inference which can also be applied to high-dimensional and non-conjugate models. We also propose a more accurate evaluation method for noise-aware posteriors. Empirically, our inference method has similar performance to existing methods in the domain where they are applicable. Outside this domain, we obtain accurate coverages on high-dimensional Bayesian linear regression and well-calibrated predictive probabilities on Bayesian logistic regression with the UCI Adult dataset.
Authors: Julie Kallini, Shikhar Murty, Christopher D. Manning, Christopher Potts, R\'obert Csord\'as
Abstract: Models that rely on subword tokenization have significant drawbacks, such as sensitivity to character-level noise like spelling errors and inconsistent compression rates across different languages and scripts. While character- or byte-level models like ByT5 attempt to address these concerns, they have not gained widespread adoption -- processing raw byte streams without tokenization results in significantly longer sequence lengths, making training and inference inefficient. This work introduces MrT5 (MergeT5), a more efficient variant of ByT5 that integrates a token deletion mechanism in its encoder to dynamically shorten the input sequence length. After processing through a fixed number of encoder layers, a learned delete gate determines which tokens are to be removed and which are to be retained for subsequent layers. MrT5 effectively "merges" critical information from deleted tokens into a more compact sequence, leveraging contextual information from the remaining tokens. In continued pre-training experiments, we find that MrT5 can achieve significant gains in inference runtime with minimal effect on performance, as measured by bits-per-byte. Additionally, with multilingual training, MrT5 adapts to the orthographic characteristics of each language, learning language-specific compression rates. Furthermore, MrT5 shows comparable accuracy to ByT5 on downstream evaluations such as XNLI, TyDi QA, and character-level tasks while reducing sequence lengths by up to 75%. Our approach presents a solution to the practical limitations of existing byte-level models.
Authors: Emanuele Marconato, S\'ebastien Lachapelle, Sebastian Weichwald, Luigi Gresele
Abstract: We analyze identifiability as a possible explanation for the ubiquity of linear properties across language models, such as the vector difference between the representations of "easy" and "easiest" being parallel to that between "lucky" and "luckiest". For this, we ask whether finding a linear property in one model implies that any model that induces the same distribution has that property, too. To answer that, we first prove an identifiability result to characterize distribution-equivalent next-token predictors, lifting a diversity requirement of previous results. Second, based on a refinement of relational linearity [Paccanaro and Hinton, 2001; Hernandez et al., 2024], we show how many notions of linearity are amenable to our analysis. Finally, we show that under suitable conditions, these linear properties either hold in all or none distribution-equivalent next-token predictors.
Authors: Fucheng Warren Zhu, Connor T. Jerzak, Adel Daoud
Abstract: Earth Observation (EO) data are increasingly used in policy analysis by enabling granular estimation of conditional average treatment effects (CATE). However, a challenge in EO-based causal inference is determining the scale of the input satellite imagery -- balancing the trade-off between capturing fine-grained individual heterogeneity in smaller images and broader contextual information in larger ones. This paper introduces Multi-Scale Representation Concatenation, a set of composable procedures that transform arbitrary single-scale EO-based CATE estimation algorithms into multi-scale ones. We benchmark the performance of Multi-Scale Representation Concatenation on a CATE estimation pipeline that combines Vision Transformer (ViT) models (which encode images) with Causal Forests (CFs) to obtain CATE estimates from those encodings. We first perform simulation studies where the causal mechanism is known, showing that our multi-scale approach captures information relevant to effect heterogeneity that single-scale ViT models fail to capture as measured by $R^2$. We then apply the multi-scale method to two randomized controlled trials (RCTs) conducted in Peru and Uganda using Landsat satellite imagery. As we do not have access to ground truth CATEs in the RCT analysis, the Rank Average Treatment Effect Ratio (RATE Ratio) measure is employed to assess performance. Results indicate that Multi-Scale Representation Concatenation improves the performance of deep learning models in EO-based CATE estimation without the complexity of designing new multi-scale architectures for a specific use case. The application of Multi-Scale Representation Concatenation could have meaningful policy benefits -- e.g., potentially increasing the impact of poverty alleviation programs without additional resource expenditure.
Authors: Robert Fonod, Haechan Cho, Hwasoo Yeo, Nikolas Geroliminis
Abstract: This paper presents a framework for extracting georeferenced vehicle trajectories from high-altitude drone imagery, addressing key challenges in urban traffic monitoring and the limitations of traditional ground-based systems. Our approach integrates several novel contributions, including a tailored object detector optimized for high-altitude bird's-eye view perspectives, a unique track stabilization method that uses detected vehicle bounding boxes as exclusion masks during image registration, and an orthophoto and master frame-based georeferencing strategy that enhances consistent alignment across multiple drone viewpoints. Additionally, our framework features robust vehicle dimension estimation and detailed road segmentation, enabling comprehensive traffic analysis. Conducted in the Songdo International Business District, South Korea, the study utilized a multi-drone experiment covering 20 intersections, capturing approximately 12TB of 4K video data over four days. The framework produced two high-quality datasets: the Songdo Traffic dataset, comprising approximately 700,000 unique vehicle trajectories, and the Songdo Vision dataset, containing over 5,000 human-annotated images with about 300,000 vehicle instances in four classes. Comparisons with high-precision sensor data from an instrumented probe vehicle highlight the accuracy and consistency of our extraction pipeline in dense urban environments. The public release of Songdo Traffic and Songdo Vision, and the complete source code for the extraction pipeline, establishes new benchmarks in data quality, reproducibility, and scalability in traffic research. Results demonstrate the potential of integrating drone technology with advanced computer vision for precise and cost-effective urban traffic monitoring, providing valuable resources for developing intelligent transportation systems and enhancing traffic management strategies.
Authors: Jannatul Chhoa, Michael Ivanitskiy, Fushuai Jiang, Shiying Li, Daniel McBride, Tom Needham, Kaiying O'Hare
Abstract: The Gromov-Wasserstein (GW) distances define a family of metrics, based on ideas from optimal transport, which enable comparisons between probability measures defined on distinct metric spaces. They are particularly useful in areas such as network analysis and geometry processing, as computation of a GW distance involves solving for registration between the objects which minimizes geometric distortion. Although GW distances have proven useful for various applications in the recent machine learning literature, it has been observed that they are inherently sensitive to outlier noise and cannot accommodate partial matching. This has been addressed by various constructions building on the GW framework; in this article, we focus specifically on a natural relaxation of the GW optimization problem, introduced by Chapel et al., which is aimed at addressing exactly these shortcomings. Our goal is to understand the theoretical properties of this relaxed optimization problem, from the viewpoint of metric geometry. While the relaxed problem fails to induce a metric, we derive precise characterizations of how it fails the axioms of non-degeneracy and triangle inequality. These observations lead us to define a novel family of distances, whose construction is inspired by the Prokhorov and Ky Fan distances, as well as by the recent work of Raghvendra et al.\ on robust versions of classical Wasserstein distance. We show that our new distances define true metrics, that they induce the same topology as the GW distances, and that they enjoy additional robustness to perturbations. These results provide a mathematically rigorous basis for using our robust partial GW distances in applications where outliers and partial matching are concerns.
Authors: Keir Adams, Kento Abeywardane, Jenna Fromer, Connor W. Coley
Abstract: Engineering molecules to exhibit precise 3D intermolecular interactions with their environment forms the basis of chemical design. In ligand-based drug design, bioisosteric analogues of known bioactive hits are often identified by virtually screening chemical libraries with shape, electrostatic, and pharmacophore similarity scoring functions. We instead hypothesize that a generative model which learns the joint distribution over 3D molecular structures and their interaction profiles may facilitate 3D interaction-aware chemical design. We specifically design ShEPhERD, an SE(3)-equivariant diffusion model which jointly diffuses/denoises 3D molecular graphs and representations of their shapes, electrostatic potential surfaces, and (directional) pharmacophores to/from Gaussian noise. Inspired by traditional ligand discovery, we compose 3D similarity scoring functions to assess ShEPhERD's ability to conditionally generate novel molecules with desired interaction profiles. We demonstrate ShEPhERD's potential for impact via exemplary drug design tasks including natural product ligand hopping, protein-blind bioactive hit diversification, and bioisosteric fragment merging.
Authors: Alexandra Butoi, Ghazal Khalighinejad, Anej Svete, Josef Valvoda, Ryan Cotterell, Brian DuSell
Abstract: Characterizing the computational power of neural network architectures in terms of formal language theory remains a crucial line of research, as it describes lower and upper bounds on the reasoning capabilities of modern AI. However, when empirically testing these bounds, existing work often leaves a discrepancy between experiments and the formal claims they are meant to support. The problem is that formal language theory pertains specifically to recognizers: machines that receive a string as input and classify whether it belongs to a language. On the other hand, it is common instead to evaluate language models on proxy tasks, e.g., language modeling or sequence-to-sequence transduction, that are similar in only an informal sense to the underlying theory. We correct this mismatch by training and evaluating neural networks directly as binary classifiers of strings, using a general method that can be applied to a wide variety of languages. As part of this, we extend an algorithm recently proposed by Sn{\ae}bjarnarson et al. (2025) for efficient length-controlled sampling of strings from regular languages. We provide results on a variety of languages across the Chomsky hierarchy for three neural architectures: a simple RNN, an LSTM, and a causally-masked transformer. We find that the RNN and LSTM often outperform the transformer, and that auxiliary training objectives such as language modeling can help, although no single objective uniformly improves performance across languages and architectures. Our contributions will facilitate theoretically sound empirical testing of language recognition claims in future work. We have released our datasets as a benchmark called FLaRe (Formal Language Recognition), along with our code.
Authors: Omri Avrahami, Or Patashnik, Ohad Fried, Egor Nemchinov, Kfir Aberman, Dani Lischinski, Daniel Cohen-Or
Abstract: Diffusion models have revolutionized the field of content synthesis and editing. Recent models have replaced the traditional UNet architecture with the Diffusion Transformer (DiT), and employed flow-matching for improved training and sampling. However, they exhibit limited generation diversity. In this work, we leverage this limitation to perform consistent image edits via selective injection of attention features. The main challenge is that, unlike the UNet-based models, DiT lacks a coarse-to-fine synthesis structure, making it unclear in which layers to perform the injection. Therefore, we propose an automatic method to identify "vital layers" within DiT, crucial for image formation, and demonstrate how these layers facilitate a range of controlled stable edits, from non-rigid modifications to object addition, using the same mechanism. Next, to enable real-image editing, we introduce an improved image inversion method for flow models. Finally, we evaluate our approach through qualitative and quantitative comparisons, along with a user study, and demonstrate its effectiveness across multiple applications. The project page is available at https://omriavrahami.com/stable-flow
Authors: Xiaoling Hu, Xiangrui Zeng, Oula Puonti, Juan Eugenio Iglesias, Bruce Fischl, Yael Balbastre
Abstract: Domain randomization through synthesis is a powerful strategy to train networks that are unbiased with respect to the domain of the input images. Randomization allows networks to see a virtually infinite range of intensities and artifacts during training, thereby minimizing overfitting to appearance and maximizing generalization to unseen data. Although powerful, this approach relies on the accurate tuning of a large set of hyperparameters that govern the probabilistic distribution of the synthesized images. Instead of manually tuning these parameters, we introduce Learn2Synth, a novel procedure in which synthesis parameters are learned using a small set of real labeled data. Unlike methods that impose constraints to align synthetic data with real data (e.g., contrastive or adversarial techniques), which risk misaligning the image and its label map, we tune an augmentation engine such that a segmentation network trained on synthetic data has optimal accuracy when applied to real data. This approach allows the training procedure to benefit from real labeled examples, without ever using these real examples to train the segmentation network, which avoids biasing the network towards the properties of the training set. Specifically, we develop parametric and nonparametric strategies to enhance synthetic images in a way that improves the performance of the segmentation network. We demonstrate the effectiveness of this learning strategy on synthetic and real-world brain scans. Code is available at: https://github.com/HuXiaoling/Learn2Synth.
Authors: Anubhav Jain, Yuya Kobayashi, Takashi Shibuya, Yuhta Takida, Nasir Memon, Julian Togelius, Yuki Mitsufuji
Abstract: Diffusion models are prone to exactly reproduce images from the training data. This exact reproduction of the training data is concerning as it can lead to copyright infringement and/or leakage of privacy-sensitive information. In this paper, we present a novel perspective on the memorization phenomenon and propose a simple yet effective approach to mitigate it. We argue that memorization occurs because of an attraction basin in the denoising process which steers the diffusion trajectory towards a memorized image. However, this can be mitigated by guiding the diffusion trajectory away from the attraction basin by not applying classifier-free guidance until an ideal transition point occurs from which classifier-free guidance is applied. This leads to the generation of non-memorized images that are high in image quality and well-aligned with the conditioning mechanism. To further improve on this, we present a new guidance technique, opposite guidance, that escapes the attraction basin sooner in the denoising process. We demonstrate the existence of attraction basins in various scenarios in which memorization occurs, and we show that our proposed approach successfully mitigates memorization.
Authors: Hanhui Wang, Yihua Zhang, Ruizheng Bai, Yue Zhao, Sijia Liu, Zhengzhong Tu
Abstract: Recent advancements in diffusion models have made generative image editing more accessible, enabling creative edits but raising ethical concerns, particularly regarding malicious edits to human portraits that threaten privacy and identity security. Existing protection methods primarily rely on adversarial perturbations to nullify edits but often fail against diverse editing requests. We propose FaceLock, a novel approach to portrait protection that optimizes adversarial perturbations to destroy or significantly alter biometric information, rendering edited outputs biometrically unrecognizable. FaceLock integrates facial recognition and visual perception into perturbation optimization to provide robust protection against various editing attempts. We also highlight flaws in commonly used evaluation metrics and reveal how they can be manipulated, emphasizing the need for reliable assessments of protection. Experiments show FaceLock outperforms baselines in defending against malicious edits and is robust against purification techniques. Ablation studies confirm its stability and broad applicability across diffusion-based editing algorithms. Our work advances biometric defense and sets the foundation for privacy-preserving practices in image editing. The code is available at: https://github.com/taco-group/FaceLock.
Authors: Jun Gao, Yongqi Li, Ziqiang Cao, Wenjie Li
Abstract: Chain-of-Thought (CoT) prompting elicits large language models (LLMs) to produce a series of intermediate reasoning steps before arriving at the final answer. However, when transitioning to vision-language models (VLMs), their text-only rationales struggle to express the fine-grained associations with the original image. In this paper, we propose an image-incorporated multimodal Chain-of-Thought, named \textbf{Interleaved-modal Chain-of-Thought (ICoT)}, which generates sequential reasoning steps consisting of paired visual and textual rationales to infer the final answer. Intuitively, the novel ICoT requires VLMs to enable the generation of fine-grained interleaved-modal content, which is hard for current VLMs to fulfill. Considering that the required visual information is usually part of the input image, we propose \textbf{Attention-driven Selection (ADS)} to realize ICoT over existing VLMs. ADS intelligently inserts regions of the input image to generate the interleaved-modal reasoning steps with ignorable additional latency. ADS relies solely on the attention map of VLMs without the need for parameterization, and therefore it is a plug-and-play strategy that can be generalized to a spectrum of VLMs. We apply ADS to realize ICoT on two popular VLMs of different architectures. Extensive evaluations of three benchmarks have shown that ICoT prompting achieves substantial performance (up to 14\%) and interpretability improvements compared to existing multimodal CoT prompting methods.
Authors: Qirui Wu, Denys Iliash, Daniel Ritchie, Manolis Savva, Angel X. Chang
Abstract: Reconstructing structured 3D scenes from RGB images using CAD objects unlocks efficient and compact scene representations that maintain compositionality and interactability. Existing works propose training-heavy methods relying on either expensive yet inaccurate real-world annotations or controllable yet monotonous synthetic data that do not generalize well to unseen objects or domains. We present Diorama, the first zero-shot open-world system that holistically models 3D scenes from single-view RGB observations without requiring end-to-end training or human annotations. We show the feasibility of our approach by decomposing the problem into subtasks and introduce robust, generalizable solutions to each: architecture reconstruction, 3D shape retrieval, object pose estimation, and scene layout optimization. We evaluate our system on both synthetic and real-world data to show we significantly outperform baselines from prior work. We also demonstrate generalization to internet images and the text-to-scene task.
Authors: Shukang Yin, Chaoyou Fu, Sirui Zhao, Yunhang Shen, Chunjiang Ge, Yan Yang, Zuwei Long, Yuhan Dai, Yongdong Luo, Haoyu Cao, Tong Xu, Xing Sun, Caifeng Shan, Ran He, Enhong Chen
Abstract: Recent years have witnessed the success of Multimodal Large Language Models (MLLMs) in the vision understanding domain. The success of these models can largely be attributed to the dominant scaling law, which states that larger parameter sizes and data volumes contribute to better performance. Notably, data scaling has mainly been powered by automatic data pipelines, which center around the self-instruction of LLMs. The paradigm has been taken for granted for quite some time, but the study of the effectiveness of scaling with these data has been neglected for a long time. In this context, this work revisits scaling with synthetic data and focuses on developing video-LLMs from a data-centric perspective. Our main study approach is fine-tuning pre-trained image-LLMs with video data and investigating learning efficiency through data scaling. Results from our preliminary experiments reveal a low learning efficiency phenomenon when simply scaling up video data samples, which, through our probing, can be ascribed to a lack of instruction diversity. Aiming at this issue, we propose a data augmentation method called Sparrow, which synthesizes video-like samples from pure text instruction data. Mixing these synthetic samples with the video data enables a more efficient training scheme. Through comprehensive experiments, we demonstrate that our proposed method achieves performance comparable to or even superior to baselines trained with many more samples. Meanwhile, we find that incorporating these synthetic samples can boost the performance of long video understanding without training with long video data. The code and data examples are available at https://github.com/VITA-MLLM/Sparrow.
Authors: Heitor R. Medeiros, Atif Belal, Srikanth Muralidharan, Eric Granger, Marco Pedersoli
Abstract: The zero-shot performance of object detectors degrades when tested on different modalities, such as infrared and depth. While recent work has explored image translation techniques to adapt detectors to new modalities, these methods are limited to a single modality and apply only to traditional detectors. Recently, vision-language detectors, such as YOLO-World and Grounding DINO, have shown promising zero-shot capabilities, however, they have not yet been adapted for other visual modalities. Traditional fine-tuning approaches compromise the zero-shot capabilities of the detectors. The visual prompt strategies commonly used for classification with vision-language models apply the same linear prompt translation to each image, making them less effective. To address these limitations, we propose ModPrompt, a visual prompt strategy to adapt vision-language detectors to new modalities without degrading zero-shot performance. In particular, an encoder-decoder visual prompt strategy is proposed, further enhanced by the integration of inference-friendly modality prompt decoupled residual, facilitating a more robust adaptation. Empirical benchmarking results show our method for modality adaptation on two vision-language detectors, YOLO-World and Grounding DINO, and on challenging infrared (LLVIP, FLIR) and depth (NYUv2) datasets, achieving performance comparable to full fine-tuning while preserving the model's zero-shot capability. Code available at: https://github.com/heitorrapela/ModPrompt.
Authors: Abulhair Saparov, Srushti Pawar, Shreyas Pimpalgaonkar, Nitish Joshi, Richard Yuanzhe Pang, Vishakh Padmakumar, Seyed Mehran Kazemi, Najoung Kim, He He
Abstract: Search is an ability foundational in many important tasks, and recent studies have shown that large language models (LLMs) struggle to perform search robustly. It is unknown whether this inability is due to a lack of data, insufficient model parameters, or fundamental limitations of the transformer architecture. In this work, we use the foundational graph connectivity problem as a testbed to generate effectively limitless high-coverage data to train small transformers and test whether they can learn to perform search. We find that, when given the right training distribution, the transformer is able to learn to search. We analyze the algorithm that the transformer has learned through a novel mechanistic interpretability technique that enables us to extract the computation graph from the trained model. We find that transformers perform search at every vertex in parallel: For each vertex in the input graph, transformers compute the set of vertices reachable from that vertex. Each layer then progressively expands these sets, allowing the model to search over a number of vertices exponential in $n_{\text{layers}}$. However, we find that as the input graph size increases, the transformer has greater difficulty in learning the task. This difficulty is not resolved even as the number of parameters is increased, suggesting that increasing model scale will not lead to robust search abilities. We also find that performing search in-context (i.e., chain-of-thought) does not resolve this inability to learn to search on larger graphs.
Authors: Fei Wu, Pablo Marquez-Neila, Hedyeh Rafi-Tarii, Raphael Sznitman
Abstract: Multi-class semantic segmentation remains a cornerstone challenge in computer vision. Yet, dataset creation remains excessively demanding in time and effort, especially for specialized domains. Active Learning (AL) mitigates this challenge by selecting data points for annotation strategically. However, existing patch-based AL methods often overlook boundary pixels critical information, essential for accurate segmentation. We present OREAL, a novel patch-based AL method designed for multi-class semantic segmentation. OREAL enhances boundary detection by employing maximum aggregation of pixel-wise uncertainty scores. Additionally, we introduce one-vs-rest entropy, a novel uncertainty score function that computes class-wise uncertainties while achieving implicit class balancing during dataset creation. Comprehensive experiments across diverse datasets and model architectures validate our hypothesis.
Authors: Alessandro Serra, Francesco Ortu, Emanuele Panizon, Lucrezia Valeriani, Lorenzo Basile, Alessio Ansuini, Diego Doimo, Alberto Cazzaniga
Abstract: Recent advances in multimodal training have significantly improved the integration of image understanding and generation within a unified model. This study investigates how vision-language models (VLMs) handle image-understanding tasks, specifically focusing on how visual information is processed and transferred to the textual domain. We compare VLMs that generate both images and text with those that output only text, highlighting key differences in information flow. We find that in models with multimodal outputs, image and text embeddings are more separated within the residual stream. Additionally, models vary in how information is exchanged from visual to textual tokens. VLMs that only output text exhibit a distributed communication pattern, where information is exchanged through multiple image tokens. In contrast, models trained for image and text generation tend to rely on a single token that acts as a narrow gate for visual information. We demonstrate that ablating this single token significantly deteriorates performance on image understanding tasks. Furthermore, modifying this token enables effective steering of the image semantics, showing that targeted, local interventions can reliably control the model's global behavior.
Authors: Siyuan Guo, Lexuan Wang, Chang Jin, Jinxian Wang, Han Peng, Huayang Shi, Wengen Li, Jihong Guan, Shuigeng Zhou
Abstract: This paper introduces M$^{3}$-20M, a large-scale Multi-Modal Molecule dataset that contains over 20 million molecules, with the data mainly being integrated from existing databases and partially generated by large language models. Designed to support AI-driven drug design and discovery, M$^{3}$-20M is 71 times more in the number of molecules than the largest existing dataset, providing an unprecedented scale that can highly benefit the training or fine-tuning of models, including large language models for drug design and discovery tasks. This dataset integrates one-dimensional SMILES, two-dimensional molecular graphs, three-dimensional molecular structures, physicochemical properties, and textual descriptions collected through web crawling and generated using GPT-3.5, offering a comprehensive view of each molecule. To demonstrate the power of M$^{3}$-20M in drug design and discovery, we conduct extensive experiments on two key tasks: molecule generation and molecular property prediction, using large language models including GLM4, GPT-3.5, GPT-4, and Llama3-8b. Our experimental results show that M$^{3}$-20M can significantly boost model performance in both tasks. Specifically, it enables the models to generate more diverse and valid molecular structures and achieve higher property prediction accuracy than existing single-modal datasets, which validates the value and potential of M$^{3}$-20M in supporting AI-driven drug design and discovery. The dataset is available at https://github.com/bz99bz/M-3.
Authors: Anubhav Jain, Yuya Kobayashi, Takashi Shibuya, Yuhta Takida, Nasir Memon, Julian Togelius, Yuki Mitsufuji
Abstract: Recent advancements in text-to-image diffusion models have brought them to the public spotlight, becoming widely accessible and embraced by everyday users. However, these models have been shown to generate harmful content such as not-safe-for-work (NSFW) images. While approaches have been proposed to erase such abstract concepts from the models, jail-breaking techniques have succeeded in bypassing such safety measures. In this paper, we propose TraSCE, an approach to guide the diffusion trajectory away from generating harmful content. Our approach is based on negative prompting, but as we show in this paper, a widely used negative prompting strategy is not a complete solution and can easily be bypassed in some corner cases. To address this issue, we first propose using a specific formulation of negative prompting instead of the widely used one. Furthermore, we introduce a localized loss-based guidance that enhances the modified negative prompting technique by steering the diffusion trajectory. We demonstrate that our proposed method achieves state-of-the-art results on various benchmarks in removing harmful content, including ones proposed by red teams, and erasing artistic styles and objects. Our proposed approach does not require any training, weight modifications, or training data (either image or prompt), making it easier for model owners to erase new concepts.
Authors: Kyle Stein, Andrew Arash Mahyari, Guillermo Francia, Eman El-Sheikh
Abstract: Backdoor attacks pose a critical threat by embedding hidden triggers into inputs, causing models to misclassify them into target labels. While extensive research has focused on mitigating these attacks in object recognition models through weight fine-tuning, much less attention has been given to detecting backdoored samples directly. Given the vast datasets used in training, manual inspection for backdoor triggers is impractical, and even state-of-the-art defense mechanisms fail to fully neutralize their impact. To address this gap, we introduce a groundbreaking method to detect unseen backdoored images during both training and inference. Leveraging the transformative success of prompt tuning in Vision Language Models (VLMs), our approach trains learnable text prompts to differentiate clean images from those with hidden backdoor triggers. Experiments demonstrate the exceptional efficacy of this method, achieving an impressive average accuracy of 86% across two renowned datasets for detecting unseen backdoor triggers, establishing a new standard in backdoor defense.
Authors: Hao Chen, Ze Wang, Xiang Li, Ximeng Sun, Fangyi Chen, Jiang Liu, Jindong Wang, Bhiksha Raj, Zicheng Liu, Emad Barsoum
Abstract: Efficient image tokenization with high compression ratios remains a critical challenge for training generative models. We present SoftVQ-VAE, a continuous image tokenizer that leverages soft categorical posteriors to aggregate multiple codewords into each latent token, substantially increasing the representation capacity of the latent space. When applied to Transformer-based architectures, our approach compresses 256x256 and 512x512 images using as few as 32 or 64 1-dimensional tokens. Not only does SoftVQ-VAE show consistent and high-quality reconstruction, more importantly, it also achieves state-of-the-art and significantly faster image generation results across different denoising-based generative models. Remarkably, SoftVQ-VAE improves inference throughput by up to 18x for generating 256x256 images and 55x for 512x512 images while achieving competitive FID scores of 1.78 and 2.21 for SiT-XL. It also improves the training efficiency of the generative models by reducing the number of training iterations by 2.3x while maintaining comparable performance. With its fully-differentiable design and semantic-rich latent space, our experiment demonstrates that SoftVQ-VAE achieves efficient tokenization without compromising generation quality, paving the way for more efficient generative models. Code and model are released.
Authors: Xiaoqing Chen, Tianwang Jia, Dongrui Wu
Abstract: An electroencephalogram (EEG) based brain-computer interface (BCI) enables direct communication between the brain and external devices. However, EEG-based BCIs face at least three major challenges in real-world applications: data scarcity and individual differences, adversarial vulnerability, and data privacy. While previous studies have addressed one or two of these issues, simultaneous accommodation of all three challenges remains challenging and unexplored. This paper fills this gap, by proposing an Aligned and Augmented Adversarial Ensemble (A3E) algorithm and integrating it into three privacy protection scenarios (centralized source-free transfer, federated source-free transfer, and source data perturbation), achieving simultaneously accurate decoding, adversarial robustness, and privacy protection of EEG-based BCIs. Experiments on three public EEG datasets demonstrated that our proposed approach outperformed over 10 classic and state-of-the-art approaches in both accuracy and robustness in all three privacy-preserving scenarios, even outperforming state-of-the-art transfer learning approaches that do not consider privacy protection at all. This is the first time that three major challenges in EEG-based BCIs can be addressed simultaneously, significantly improving the practicalness of EEG decoding in real-world BCIs.
Authors: Jiale Cheng, Xiao Liu, Cunxiang Wang, Xiaotao Gu, Yida Lu, Dan Zhang, Yuxiao Dong, Jie Tang, Hongning Wang, Minlie Huang
Abstract: Instruction-following is a fundamental capability of language models, requiring the model to recognize even the most subtle requirements in the instructions and accurately reflect them in its output. Such an ability is well-suited for and often optimized by preference learning. However, existing methods often directly sample multiple independent responses from the model when creating preference pairs. Such practice can introduce content variations irrelevant to whether the instruction is precisely followed (e.g., different expressions about the same semantic), interfering with the goal of teaching models to recognize the key differences that lead to improved instruction following. In light of this, we introduce SPaR, a self-play framework integrating tree-search self-refinement to yield valid and comparable preference pairs free from distractions. By playing against itself, an LLM employs a tree-search strategy to refine its previous responses with respect to the instruction while minimizing unnecessary variations. Our experiments show that a LLaMA3-8B model, trained over three iterations guided by SPaR, surpasses GPT-4-Turbo on the IFEval benchmark without losing general capabilities. Furthermore, SPaR demonstrates promising scalability, greatly enhancing models like GLM-4-9B and LLaMA3-70B. We also identify how inference scaling in tree search would impact model performance. Our code and data are publicly available at https://github.com/thu-coai/SPaR.
Authors: Stanislas Ducotterd, Sebastian Neumayer, Michael Unser
Abstract: We aim at the solution of inverse problems in imaging, by combining a penalized sparse representation of image patches with an unconstrained smooth one. This allows for a straightforward interpretation of the reconstruction. We formulate the optimization as a bilevel problem. The inner problem deploys classical algorithms while the outer problem optimizes the dictionary and the regularizer parameters through supervised learning. The process is carried out via implicit differentiation and gradient-based optimization. We evaluate our method for denoising, super-resolution, and compressed-sensing magnetic-resonance imaging. We compare it to other classical models as well as deep-learning-based methods and show that it always outperforms the former and also the latter in some instances.
Authors: Mahdi Saberi, Chi Zhang, Mehmet Akcakaya
Abstract: Deep learning (DL) methods, especially those based on physics-driven DL, have become the state-of-the-art for reconstructing sub-sampled magnetic resonance imaging (MRI) data. However, studies have shown that these methods are susceptible to small adversarial input perturbations, or attacks, resulting in major distortions in the output images. Various strategies have been proposed to reduce the effects of these attacks, but they require retraining and may lower reconstruction quality for non-perturbed/clean inputs. In this work, we propose a novel approach for mitigating adversarial attacks on MRI reconstruction models without any retraining. Our framework is based on the idea of cyclic measurement consistency. The output of the model is mapped to another set of MRI measurements for a different sub-sampling pattern, and this synthesized data is reconstructed with the same model. Intuitively, without an attack, the second reconstruction is expected to be consistent with the first, while with an attack, disruptions are present. A novel objective function is devised based on this idea, which is minimized within a small ball around the attack input for mitigation. Experimental results show that our method substantially reduces the impact of adversarial perturbations across different datasets, attack types/strengths and PD-DL networks, and qualitatively and quantitatively outperforms conventional mitigation methods that involve retraining. Finally, we extend our mitigation method to two important practical scenarios: a blind setup, where the attack strength or algorithm is not known to the end user; and an adaptive attack setup, where the attacker has full knowledge of the defense strategy. Our approach remains effective in both cases.
Authors: Zongxia Li, Xiyang Wu, Hongyang Du, Huy Nghiem, Guangyao Shi
Abstract: Multimodal Vision Language Models (VLMs) have emerged as a transformative technology at the intersection of computer vision and natural language processing, enabling machines to perceive and reason about the world through both visual and textual modalities. For example, models such as CLIP, Claude, and GPT-4V demonstrate strong reasoning and understanding abilities on visual and textual data and beat classical single modality vision models on zero-shot classification. Despite their rapid advancements in research and growing popularity in applications, a comprehensive survey of existing studies on VLMs is notably lacking, particularly for researchers aiming to leverage VLMs in their specific domains. To this end, we provide a systematic overview of VLMs in the following aspects: model information of the major VLMs developed over the past five years (2019-2024); the main architectures and training methods of these VLMs; summary and categorization of the popular benchmarks and evaluation metrics of VLMs; the applications of VLMs including embodied agents, robotics, and video generation; the challenges and issues faced by current VLMs such as hallucination, fairness, and safety. Detailed collections including papers and model repository links are listed in https://github.com/zli12321/Vision-Language-Models-Overview.
URLs: https://github.com/zli12321/Vision-Language-Models-Overview.
Authors: Yooseop Lee, Suin Kim, Yohan Jo
Abstract: In designing multiple-choice questions (MCQs) in education, creating plausible distractors is crucial for identifying students' misconceptions and gaps in knowledge and accurately assessing their understanding. However, prior studies on distractor generation have not paid sufficient attention to enhancing the difficulty of distractors, resulting in reduced effectiveness of MCQs. This study presents a pipeline for training a model to generate distractors that are more likely to be selected by students. First, we train a pairwise ranker to reason about students' misconceptions and assess the relative plausibility of two distractors. Using this model, we create a dataset of pairwise distractor ranks and then train a distractor generator via Direct Preference Optimization (DPO) to generate more plausible distractors. Experiments on computer science subjects (Python, DB, MLDL) demonstrate that our pairwise ranker effectively identifies students' potential misunderstandings and achieves ranking accuracy comparable to human experts. Furthermore, our distractor generator outperforms several baselines in generating plausible distractors and produces questions with a higher item discrimination index (DI).
Authors: Kavitha Viswanathan, Shashwat Pathak, Piyush Bharambe, Harsh Choudhary, Amit Sethi
Abstract: The tradeoff between reconstruction quality and compute required for video super-resolution (VSR) remains a formidable challenge in its adoption for deployment on resource-constrained edge devices. While transformer-based VSR models have set new benchmarks for reconstruction quality in recent years, these require substantial computational resources. On the other hand, lightweight models that have been introduced even recently struggle to deliver state-of-the-art reconstruction. We propose a novel lightweight and parameter-efficient neural architecture for VSR that achieves state-of-the-art reconstruction accuracy with just 2.3 million parameters. Our model enhances information utilization based on several architectural attributes. Firstly, it uses 2D wavelet decompositions strategically interlayered with learnable convolutional layers to utilize the inductive prior of spatial sparsity of edges in visual data. Secondly, it uses a single memory tensor to capture inter-frame temporal information while avoiding the computational cost of previous memory-based schemes. Thirdly, it uses residual deformable convolutions for implicit inter-frame object alignment that improve upon deformable convolutions by enhancing spatial information in inter-frame feature differences. Architectural insights from our model can pave the way for real-time VSR on the edge, such as display devices for streaming data.
Authors: Tansheng Zhu, Hongyu Zhou, Ke Jin, Xusheng Xu, Qiufan Yuan, Lijie Ji
Abstract: Bayesian optimization is highly effective for optimizing expensive-to-evaluate black-box functions, but it faces significant computational challenges due to the high computational complexity of Gaussian processes, which results in a total time complexity that is quartic with respect to the number of iterations. To address this limitation, we propose the Bayesian Optimization by Kernel regression and density-based Exploration (BOKE) algorithm. BOKE uses kernel regression for efficient function approximation, kernel density for exploration, and integrates them into the confidence bound criteria to guide the optimization process, thus reducing computational costs to quadratic. Our theoretical analysis rigorously establishes the global convergence of BOKE and ensures its robustness in noisy settings. Through extensive numerical experiments on both synthetic and real-world optimization tasks, we demonstrate that BOKE not only performs competitively compared to Gaussian process-based methods but also exhibits superior computational efficiency. These results highlight BOKE's effectiveness in resource-constrained environments, providing a practical approach for optimization problems in engineering applications.
Authors: Lingze Zhang, Ellie Pavlick
Abstract: An increasingly common practice is to train large language models (LLMs) using synthetic data. Often this synthetic data is produced by the same or similar LLMs as those it is being used to train. This raises the question of whether the synthetic data might in fact exacerbate certain "blindspots" by reinforcing heuristics that the LLM already encodes. In this paper, we conduct simulated experiments on the natural language inference (NLI) task with Llama-2-7B-hf models. We use MultiNLI as the general task and HANS, a targeted evaluation set designed to measure the presence of specific heuristic strategies for NLI, as our "blindspot" task. Our goal is to determine whether performance disparities between the general and blind spot tasks emerge. Our results indicate that synthetic data does not reinforce blindspots in the way we expected. Specifically, we see that, while fine-tuning with synthetic data doesn't necessarily reduce the use of the heuristic, it also does not make it worse as we hypothesized.
Authors: Ya-Chi Chu, Wenzhi Gao, Yinyu Ye, Madeleine Udell
Abstract: This paper investigates the convergence properties of the hypergradient descent method (HDM), a 25-year-old heuristic originally proposed for adaptive stepsize selection in stochastic first-order methods. We provide the first rigorous convergence analysis of HDM using the online learning framework of [Gao24] and apply this analysis to develop new state-of-the-art adaptive gradient methods with empirical and theoretical support. Notably, HDM automatically identifies the optimal stepsize for the local optimization landscape and achieves local superlinear convergence. Our analysis explains the instability of HDM reported in the literature and proposes efficient strategies to address it. We also develop two HDM variants with heavy-ball and Nesterov momentum. Experiments on deterministic convex problems show HDM with heavy-ball momentum (HDM-HB) exhibits robust performance and significantly outperforms other adaptive first-order methods. Moreover, HDM-HB often matches the performance of L-BFGS, an efficient and practical quasi-Newton method, using less memory and cheaper iterations.
Authors: Alessandro Favero, Antonio Sclocchi, Francesco Cagnetta, Pascal Frossard, Matthieu Wyart
Abstract: Natural data is often organized as a hierarchical composition of features. How many samples do generative models need in order to learn the composition rules, so as to produce a combinatorially large number of novel data? What signal in the data is exploited to learn those rules? We investigate these questions in the context of diffusion models both theoretically and empirically. Theoretically, we consider simple probabilistic context-free grammars - tree-like graphical models used to represent the hierarchical and compositional structure of data such as language and images. We demonstrate that diffusion models learn the grammar's composition rules with the sample complexity required for clustering features with statistically similar context, a process similar to the word2vec algorithm. However, this clustering emerges hierarchically: higher-level features associated with longer contexts require more data to be identified. This mechanism leads to a sample complexity that scales polynomially with the said context size. As a result, diffusion models trained on an intermediate dataset size generate data coherent up to a certain scale, but that lacks global coherence. We test these predictions in different domains, and find remarkable agreement: both generated texts and images achieve progressively larger coherence lengths as the training time or dataset size grows. We discuss connections between the hierarchical clustering mechanism we introduce here and the renormalization group in physics.
Authors: Ehab Ghannoum, Mohammad Ghafari
Abstract: Deep learning models have gained popularity for conducting various tasks involving source code. However, their black-box nature raises concerns about potential risks. One such risk is a poisoning attack, where an attacker intentionally contaminates the training set with malicious samples to mislead the model's predictions in specific scenarios. To protect source code models from poisoning attacks, we introduce CodeGarrison (CG), a hybrid deep-learning model that relies on code embeddings to identify poisoned code samples. We evaluated CG against the state-of-the-art technique ONION for detecting poisoned samples generated by DAMP, MHM, ALERT, as well as a novel poisoning technique named CodeFooler. Results showed that CG significantly outperformed ONION with an accuracy of 93.5%. We also tested CG's robustness against unknown attacks, achieving an average accuracy of 85.6% in identifying poisoned samples across the four attacks mentioned above.
Authors: George Drayson, Vasileios Lampos
Abstract: As Large Language Models (LLMs) become increasingly prevalent, their generated outputs are proliferating across the web, risking a future where machine-generated content dilutes human-authored text. Since online data is the primary resource for LLM pre-training, subsequent models could be trained on an unknown portion of synthetic samples. This will lead to model collapse, a degenerative process whereby LLMs reinforce their own errors, and ultimately yield a declining performance. In this study, we investigate the impact of decoding strategy on model collapse, analysing the characteristics of text at each model generation, the similarity to human references, and the resulting model performance. Using the decoding strategies that lead to the most significant degradation, we evaluate model collapse in more realistic scenarios where the origin of the data (human or synthetic) is unknown. We train a machine-generated text detector and propose an importance sampling approach to alleviate model collapse. Our method is validated on two LLM variants (GPT-2 and SmolLM2) on the open-ended text generation task. We demonstrate that it can not only prevent model collapse but also improve performance when sufficient human-authored samples are present.
Authors: Tian Yu Liu, Alessandro Achille, Matthew Trager, Aditya Golatkar, Luca Zancato, Stefano Soatto
Abstract: Providing Large Language Models with relevant contextual knowledge at inference time has been shown to greatly improve the quality of their generations. This is often achieved by prepending informative passages of text, or 'contexts', retrieved from external knowledge bases to their input. However, processing additional contexts online incurs significant computation costs that scale with their length. State Space Models (SSMs) offer a promising solution by allowing a database of contexts to be mapped onto fixed-dimensional states from which to start the generation. A key challenge arises when attempting to leverage information present across multiple contexts, since there is no straightforward way to condition generation on multiple independent states in existing SSMs. To address this, we leverage a simple mathematical relation derived from SSM dynamics to compose multiple states into one that efficiently approximates the effect of concatenating raw context tokens. Since the temporal ordering of contexts can often be uninformative, we enforce permutation-invariance by efficiently averaging states obtained via our composition algorithm across all possible context orderings. We evaluate our resulting method on WikiText and MSMARCO in both zero-shot and fine-tuned settings, and show that we can match the strongest performing baseline while enjoying on average 5.4x speedup.
Authors: Taishi Nakamura, Takuya Akiba, Kazuki Fujii, Yusuke Oda, Rio Yokota, Jun Suzuki
Abstract: The Mixture of Experts (MoE) architecture reduces the training and inference cost significantly compared to a dense model of equivalent capacity. Upcycling is an approach that initializes and trains an MoE model using a pre-trained dense model. While upcycling leads to initial performance gains, the training progresses slower than when trained from scratch, leading to suboptimal performance in the long term. We propose Drop-Upcycling - a method that effectively addresses this problem. Drop-Upcycling combines two seemingly contradictory approaches: utilizing the knowledge of pre-trained dense models while statistically re-initializing some parts of the weights. This approach strategically promotes expert specialization, significantly enhancing the MoE model's efficiency in knowledge acquisition. Extensive large-scale experiments demonstrate that Drop-Upcycling significantly outperforms previous MoE construction methods in the long term, specifically when training on hundreds of billions of tokens or more. As a result, our MoE model with 5.9B active parameters achieves comparable performance to a 13B dense model in the same model family, while requiring approximately 1/4 of the training FLOPs. All experimental resources, including source code, training data, model checkpoints and logs, are publicly available to promote reproducibility and future research on MoE.
Authors: Samet Demir, Zafer Dogan
Abstract: In this work, we study the training and generalization performance of two-layer neural networks (NNs) after one gradient descent step under structured data modeled by Gaussian mixtures. While previous research has extensively analyzed this model under isotropic data assumption, such simplifications overlook the complexities inherent in real-world datasets. Our work addresses this limitation by analyzing two-layer NNs under Gaussian mixture data assumption in the asymptotically proportional limit, where the input dimension, number of hidden neurons, and sample size grow with finite ratios. We characterize the training and generalization errors by leveraging recent advancements in Gaussian universality. Specifically, we prove that a high-order polynomial model performs equivalent to the nonlinear neural networks under certain conditions. The degree of the equivalent model is intricately linked to both the "data spread" and the learning rate employed during one gradient step. Through extensive simulations, we demonstrate the equivalence between the original model and its polynomial counterpart across various regression and classification tasks. Additionally, we explore how different properties of Gaussian mixtures affect learning outcomes. Finally, we illustrate experimental results on Fashion-MNIST classification, indicating that our findings can translate to realistic data.
Authors: Sueda Taner, Ziyi Wang, Christoph Studer
Abstract: We introduce a novel class of regularization functions, called Cauchy-Schwarz (CS) regularizers, which can be designed to induce a wide range of properties in solution vectors of optimization problems. To demonstrate the versatility of CS regularizers, we derive regularization functions that promote discrete-valued vectors, eigenvectors of a given matrix, and orthogonal matrices. The resulting CS regularizers are simple, differentiable, and can be free of spurious stationary points, making them suitable for gradient-based solvers and large-scale optimization problems. In addition, CS regularizers automatically adapt to the appropriate scale, which is, for example, beneficial when discretizing the weights of neural networks. To demonstrate the efficacy of CS regularizers, we provide results for solving underdetermined systems of linear equations and weight quantization in neural networks. Furthermore, we discuss specializations, variations, and generalizations, which lead to an even broader class of new and possibly more powerful regularizers.
Authors: Zhihao Zhan, Wang Pang, Xiang Zhu, Yechao Bai
Abstract: We present a generic video super-resolution algorithm in this paper, based on the Diffusion Posterior Sampling framework with an unconditional video generation model in latent space. The video generation model, a diffusion transformer, functions as a space-time model. We argue that a powerful model, which learns the physics of the real world, can easily handle various kinds of motion patterns as prior knowledge, thus eliminating the need for explicit estimation of optical flows or motion parameters for pixel alignment. Furthermore, a single instance of the proposed video diffusion transformer model can adapt to different sampling conditions without re-training. Empirical results on synthetic and real-world datasets demonstrate that our method has strong capabilities to address video super-resolution challenges.
Authors: Yixin Su, Wei Jiang, Fangquan Lin, Cheng Yang, Sarah M. Erfani, Junhao Gan, Yunxiang Zhao, Ruixuan Li, Rui Zhang
Abstract: In recommender systems, the patterns of user behaviors (e.g., purchase, click) may vary greatly in different contexts (e.g., time and location). This is because user behavior is jointly determined by two types of factors: intrinsic factors, which reflect consistent user preference, and extrinsic factors, which reflect external incentives that may vary in different contexts. Differentiating between intrinsic and extrinsic factors helps learn user behaviors better. However, existing studies have only considered differentiating them from a single, pre-defined context (e.g., time or location), ignoring the fact that a user's extrinsic factors may be influenced by the interplay of various contexts at the same time. In this paper, we propose the Intrinsic-Extrinsic Disentangled Recommendation (IEDR) model, a generic framework that differentiates intrinsic from extrinsic factors considering various contexts simultaneously, enabling more accurate differentiation of factors and hence the improvement of recommendation accuracy. IEDR contains a context-invariant contrastive learning component to capture intrinsic factors, and a disentanglement component to extract extrinsic factors under the interplay of various contexts. The two components work together to achieve effective factor learning. Extensive experiments on real-world datasets demonstrate IEDR's effectiveness in learning disentangled factors and significantly improving recommendation accuracy by up to 4% in NDCG.
Authors: Jacqueline R. M. A. Maasch, Alihan H\"uy\"uk, Xinnuo Xu, Aditya V. Nori, Javier Gonzalez
Abstract: Causal reasoning and compositional reasoning are two core aspirations in generative AI. Measuring the extent of these behaviors requires principled evaluation methods. We explore a unified perspective that considers both behaviors simultaneously, termed compositional causal reasoning (CCR): the ability to infer how causal measures compose and, equivalently, how causal quantities propagate through graphs. We instantiate a framework for the systematic evaluation of CCR for the average treatment effect and the probability of necessity and sufficiency. As proof of concept, we demonstrate the design of CCR tasks for language models in the LLama, Phi, and GPT families. On a math word problem, our framework revealed a range of taxonomically distinct error patterns. Additionally, CCR errors increased with the complexity of causal paths for all models except o1.
Authors: Yijie Xu, Aiwei Liu, Xuming Hu, Lijie Wen, Hui Xiong
Abstract: As open-source large language models (LLMs) like Llama3 become more capable, it is crucial to develop watermarking techniques to detect their potential misuse. Existing watermarking methods either add watermarks during LLM inference, which is unsuitable for open-source LLMs, or primarily target classification LLMs rather than recent generative LLMs. Adapting these watermarks to open-source LLMs for misuse detection remains an open challenge. This work defines two misuse scenarios for open-source LLMs: intellectual property (IP) violation and LLM Usage Violation. Then, we explore the application of inference-time watermark distillation and backdoor watermarking in these contexts. We propose comprehensive evaluation methods to assess the impact of various real-world further fine-tuning scenarios on watermarks and the effect of these watermarks on LLM performance. Our experiments reveal that backdoor watermarking could effectively detect IP Violation, while inference-time watermark distillation is applicable in both scenarios but less robust to further fine-tuning and has a more significant impact on LLM performance compared to backdoor watermarking. Exploring more advanced watermarking methods for open-source LLMs to detect their misuse should be an important future direction.
Authors: Xubin Wang, Zhiqing Tang, Jianxiong Guo, Tianhui Meng, Chenhao Wang, Tian Wang, Weijia Jia
Abstract: The rapid advancement of artificial intelligence (AI) technologies has led to an increasing deployment of AI models on edge and terminal devices, driven by the proliferation of the Internet of Things (IoT) and the need for real-time data processing. This survey comprehensively explores the current state, technical challenges, and future trends of on-device AI models. We define on-device AI models as those designed to perform local data processing and inference, emphasizing their characteristics such as real-time performance, resource constraints, and enhanced data privacy. The survey is structured around key themes, including the fundamental concepts of AI models, application scenarios across various domains, and the technical challenges faced in edge environments. We also discuss optimization and implementation strategies, such as data preprocessing, model compression, and hardware acceleration, which are essential for effective deployment. Furthermore, we examine the impact of emerging technologies, including edge computing and foundation models, on the evolution of on-device AI models. By providing a structured overview of the challenges, solutions, and future directions, this survey aims to facilitate further research and application of on-device AI, ultimately contributing to the advancement of intelligent systems in everyday life.
Authors: Edwin Hamel-De le Court, Francesco Belardinelli, Alex W. Goodall
Abstract: In real-life scenarios, a Reinforcement Learning (RL) agent aiming to maximise their reward, must often also behave in a safe manner, including at training time. Thus, much attention in recent years has been given to Safe RL, where an agent aims to learn an optimal policy among all policies that satisfy a given safety constraint. However, strict safety guarantees are often provided through approaches based on linear programming, and thus have limited scaling. In this paper we present a new, scalable method, which enjoys strict formal guarantees for Safe RL, in the case where the safety dynamics of the Markov Decision Process (MDP) are known, and safety is defined as an undiscounted probabilistic avoidance property. Our approach is based on state-augmentation of the MDP, and on the design of a shield that restricts the actions available to the agent. We show that our approach provides a strict formal safety guarantee that the agent stays safe at training and test time. Furthermore, we demonstrate that our approach is viable in practice through experimental evaluation.
Authors: Steeven Janny, Herv\'e Poirier, Leonid Antsfeld, Guillaume Bono, Gianluca Monaci, Boris Chidlovskii, Francesco Giuliari, Alessio Del Bue, Christian Wolf
Abstract: Progress in Embodied AI has made it possible for end-to-end-trained agents to navigate in photo-realistic environments with high-level reasoning and zero-shot or language-conditioned behavior, but benchmarks are still dominated by simulation. In this work, we focus on the fine-grained behavior of fast-moving real robots and present a large-scale experimental study involving \numepisodes{} navigation episodes in a real environment with a physical robot, where we analyze the type of reasoning emerging from end-to-end training. In particular, we study the presence of realistic dynamics which the agent learned for open-loop forecasting, and their interplay with sensing. We analyze the way the agent uses latent memory to hold elements of the scene structure and information gathered during exploration. We probe the planning capabilities of the agent, and find in its memory evidence for somewhat precise plans over a limited horizon. Furthermore, we show in a post-hoc analysis that the value function learned by the agent relates to long-term planning. Put together, our experiments paint a new picture on how using tools from computer vision and sequential decision making have led to new capabilities in robotics and control. An interactive tool is available at europe.naverlabs.com/research/publications/reasoning-in-visual-navigation-of-end-to-end-trained-agents.
Authors: Bin Yang, Yuxuan Liang, Chenjuan Guo, Christian S. Jensen
Abstract: Time series data captures properties that change over time. Such data occurs widely, ranging from the scientific and medical domains to the industrial and environmental domains. When the properties in time series exhibit spatial variations, we often call the data spatio-temporal. As part of the continued digitalization of processes throughout society, increasingly large volumes of time series and spatio-temporal data are available. In this tutorial, we focus on data-driven decision making with such data, e.g., enabling greener and more efficient transportation based on traffic time series forecasting. The tutorial adopts the holistic paradigm of "data-governance-analytics-decision." We first introduce the data foundation of time series and spatio-temporal data, which is often heterogeneous. Next, we discuss data governance methods that aim to improve data quality. We then cover data analytics, focusing on five desired characteristics: automation, robustness, generality, explainability, and resource efficiency. We finally cover data-driven decision making strategies and briefly discuss promising research directions. We hope that the tutorial will serve as a primary resource for researchers and practitioners who are interested in value creation from time series and spatio-temporal data.
Authors: Aviad Barzilai, Yotam Gigi, Amr Helmy, Vered Silverman, Yehonathan Refael, Bolous Jaber, Tomer Shekel, George Leifman, Genady Beryozkin
Abstract: Foundation models have had a significant impact across various AI applications, enabling use cases that were previously impossible. Contrastive Visual Language Models (VLMs), in particular, have outperformed other techniques in many tasks. However, their prevalence in remote sensing (RS) is still limited, due to the scarcity of diverse remote-sensing visual-language datasets. In this work we introduce two novel image-caption datasets for training of remote sensing foundation models. The first dataset pairs aerial and satellite imagery with captions generated by Gemini using landmarks extracted from Google Maps. The second dataset utilizes public web images and their corresponding alt-text, filtered for the remote sensing domain, resulting in a diverse dataset with greater breadth in image styles and subject matter. These datasets are used to pre-train the MaMMUT~\citep{kuo2023mammutsimplearchitecturejoint} VLM architecture, resulting in state-of-the-art generalization performance in zero-shot cross-modal retrieval on well-known public benchmarks. Finally, we present our ongoing research to distill image-level knowledge gained in the VLM contrastive training procedure to enhance the model's localization ability. Specifically, we iteratively generate pseudo-labels for image regions based on the model's attention maps and use these labels for further training. To mitigate noisy attention maps and create robust segmentation masks, we introduce a novel attention-pooling mechanism called the Smooth-Attention-Operation.
Authors: Janis Zenkner, Tobias Sesterhenn, Christian Bartelt
Abstract: Task decomposition is a fundamental mechanism in program synthesis, enabling complex problems to be broken down into manageable subtasks. ExeDec, a state-of-the-art program synthesis framework, employs this approach by combining a Subgoal Model for decomposition and a Synthesizer Model for program generation to facilitate compositional generalization. In this work, we develop REGISM, an adaptation of ExeDec that removes decomposition guidance and relies solely on iterative execution-driven synthesis. By comparing these two exemplary approaches-ExeDec, which leverages task decomposition, and REGISM, which does not-we investigate the interplay between task decomposition and program generation. Our findings indicate that ExeDec exhibits significant advantages in length generalization and concept composition tasks, likely due to its explicit decomposition strategies. At the same time, REGISM frequently matches or surpasses ExeDec's performance across various scenarios, with its solutions often aligning more closely with ground truth decompositions. These observations highlight the importance of repeated execution-guided synthesis in driving task-solving performance, even within frameworks that incorporate explicit decomposition strategies. Our analysis suggests that task decomposition approaches like ExeDec hold significant potential for advancing program synthesis, though further work is needed to clarify when and why these strategies are most effective.
Authors: Stefan Sylvius Wagner, Stefan Harmeling
Abstract: Object-centric learning is fundamental to human vision and crucial for models requiring complex reasoning. Traditional approaches rely on slot-based bottlenecks to learn object properties explicitly, while recent self-supervised vision models like DINO have shown emergent object understanding. However, DINO representations primarily capture global scene features, often confounding individual object attributes. We investigate the effectiveness of DINO representations and slot-based methods for multi-object instance retrieval. Our findings reveal that DINO representations excel at capturing global object attributes such as object shape and size, but struggle with object-level details like colour, whereas slot-based representations struggle at both global and object-level understanding. To address this, we propose a method that combines global and local features by augmenting DINO representations with object-centric latent vectors from a Variational Autoencoder trained on segmented image patches that are extracted from the DINO features. This approach improves multi-object instance retrieval performance, bridging the gap between global scene understanding and fine-grained object representation without requiring full model retraining.
Authors: Jiawei Zhang, Ziyuan Liu, Leon Yan, Gen Li, Yuantao Gu
Abstract: Diffusion models have demonstrated remarkable performance in modeling complex data priors, catalyzing their widespread adoption in solving various inverse problems. However, the inherently iterative nature of diffusion-based inverse algorithms often requires hundreds to thousands of steps, with performance degradation occurring under fewer steps which limits their practical applicability. While high-order diffusion ODE solvers have been extensively explored for efficient diffusion sampling without observations, their application to inverse problems remains underexplored due to the diverse forms of inverse algorithms and their need for repeated trajectory correction based on observations. To address this gap, we first introduce a canonical form that decomposes existing diffusion-based inverse algorithms into three modules to unify their analysis. Inspired by the linear subspace search strategy in the design of high-order diffusion ODE solvers, we propose the Learnable Linear Extrapolation (LLE) method, a lightweight approach that universally enhances the performance of any diffusion-based inverse algorithm that fits the proposed canonical form. Extensive experiments demonstrate consistent improvements of the proposed LLE method across multiple algorithms and tasks, indicating its potential for more efficient solutions and boosted performance of diffusion-based inverse algorithms with limited steps. Codes for reproducing our experiments are available at https://github.com/weigerzan/LLE_inverse_problem.
Authors: Shailesh Lal, Suvajit Majumder, Evgeny Sobko
Abstract: We introduce a novel machine learning based framework for discovering integrable models. Our approach first employs a synchronized ensemble of neural networks to find high-precision numerical solution to the Yang-Baxter equation within a specified class. Then, using an auxiliary system of algebraic equations, [Q_2, Q_3] = 0, and the numerical value of the Hamiltonian obtained via deep learning as a seed, we reconstruct the entire Hamiltonian family, forming an algebraic variety. We illustrate our presentation with three- and four-dimensional spin chains of difference form with local interactions. Remarkably, all discovered Hamiltonian families form rational varieties.