Authors: Xinhua Wu, Haoyu He, Yanchao Wang, Qi Wang
Abstract: Ubiquitous mobile devices are generating vast amounts of location-based service data that reveal how individuals navigate and utilize urban spaces in detail. In this study, we utilize these extensive, unlabeled sequences of user trajectories to develop a foundation model for understanding urban space and human mobility. We introduce the \textbf{P}retrained \textbf{M}obility \textbf{T}ransformer (PMT), which leverages the transformer architecture to process user trajectories in an autoregressive manner, converting geographical areas into tokens and embedding spatial and temporal information within these representations. Experiments conducted in three U.S. metropolitan areas over a two-month period demonstrate PMT's ability to capture underlying geographic and socio-demographic characteristics of regions. The proposed PMT excels across various downstream tasks, including next-location prediction, trajectory imputation, and trajectory generation. These results support PMT's capability and effectiveness in decoding complex patterns of human mobility, offering new insights into urban spatial functionality and individual mobility preferences.
Authors: Grant Norman, Jacqueline Wentz, Hemanth Kolla, Kurt Maute, Alireza Doostan
Abstract: Throughout many fields, practitioners often rely on differential equations to model systems. Yet, for many applications, the theoretical derivation of such equations and/or accurate resolution of their solutions may be intractable. Instead, recently developed methods, including those based on parameter estimation, operator subset selection, and neural networks, allow for the data-driven discovery of both ordinary and partial differential equations (PDEs), on a spectrum of interpretability. The success of these strategies is often contingent upon the correct identification of representative equations from noisy observations of state variables and, as importantly and intertwined with that, the mathematical strategies utilized to enforce those equations. Specifically, the latter has been commonly addressed via unconstrained optimization strategies. Representing the PDE as a neural network, we propose to discover the PDE by solving a constrained optimization problem and using an intermediate state representation similar to a Physics-Informed Neural Network (PINN). The objective function of this constrained optimization problem promotes matching the data, while the constraints require that the PDE is satisfied at several spatial collocation points. We present a penalty method and a widely used trust-region barrier method to solve this constrained optimization problem, and we compare these methods on numerical examples. Our results on the Burgers' and the Korteweg-De Vreis equations demonstrate that the latter constrained method outperforms the penalty method, particularly for higher noise levels or fewer collocation points. For both methods, we solve these discovered neural network PDEs with classical methods, such as finite difference methods, as opposed to PINNs-type methods relying on automatic differentiation. We briefly highlight other small, yet crucial, implementation details.
Authors: Yinan Wang, M. Giselle Fern\'andez-Godino, Nipun Gunawardena, Donald D. Lucas, Xiaowei Yue
Abstract: Industrial accidents, chemical spills, and structural fires can release large amounts of harmful materials that disperse into urban atmospheres and impact populated areas. Computer models are typically used to predict the transport of toxic plumes by solving fluid dynamical equations. However, these models can be computationally expensive due to the need for many grid cells to simulate turbulent flow and resolve individual buildings and streets. In emergency response situations, alternative methods are needed that can run quickly and adequately capture important spatiotemporal features. Here, we present a novel deep learning model called ST-GasNet that was inspired by the mathematical equations that govern the behavior of plumes as they disperse through the atmosphere. ST-GasNet learns the spatiotemporal dependencies from a limited set of temporal sequences of ground-level toxic urban plumes generated by a high-resolution large eddy simulation model. On independent sequences, ST-GasNet accurately predicts the late-time spatiotemporal evolution, given the early-time behavior as an input, even for cases when a building splits a large plume into smaller plumes. By incorporating large-scale wind boundary condition information, ST-GasNet achieves a prediction accuracy of at least 90% on test data for the entire prediction period.
Authors: Seyd Teymoor Seydi
Abstract: This paper presents a comprehensive survey of 18 distinct polynomials and their potential applications in Kolmogorov-Arnold Network (KAN) models as an alternative to traditional spline-based methods. The polynomials are classified into various groups based on their mathematical properties, such as orthogonal polynomials, hypergeometric polynomials, q-polynomials, Fibonacci-related polynomials, combinatorial polynomials, and number-theoretic polynomials. The study aims to investigate the suitability of these polynomials as basis functions in KAN models for complex tasks like handwritten digit classification on the MNIST dataset. The performance metrics of the KAN models, including overall accuracy, Kappa, and F1 score, are evaluated and compared. The Gottlieb-KAN model achieves the highest performance across all metrics, suggesting its potential as a suitable choice for the given task. However, further analysis and tuning of these polynomials on more complex datasets are necessary to fully understand their capabilities in KAN models. The source code for the implementation of these KAN models is available at https://github.com/seydi1370/Basis_Functions .
Authors: Kazuki Sakamoto, Connor T. Jerzak, Adel Daoud
Abstract: Earth observation data such as satellite imagery can, when combined with machine learning, have profound impacts on our understanding of the geography of poverty through the prediction of living conditions, especially where government-derived economic indicators are either unavailable or potentially untrustworthy. Recent work has progressed in using EO data not only to predict spatial economic outcomes, but also to explore cause and effect, an understanding which is critical for downstream policy analysis. In this review, we first document the growth of interest in EO-ML analyses in the causal space. We then trace the relationship between spatial statistics and EO-ML methods before discussing the four ways in which EO data has been used in causal ML pipelines -- (1.) poverty outcome imputation for downstream causal analysis, (2.) EO image deconfounding, (3.) EO-based treatment effect heterogeneity, and (4.) EO-based transportability analysis. We conclude by providing a workflow for how researchers can incorporate EO data in causal ML analysis going forward.
Authors: Siavash Golkar, Alberto Bietti, Mariel Pettee, Michael Eickenberg, Miles Cranmer, Keiya Hirashima, Geraud Krawezik, Nicholas Lourie, Michael McCabe, Rudy Morel, Ruben Ohana, Liam Holden Parker, Bruno R\'egaldo-Saint Blancard, Kyunghyun Cho, Shirley Ho
Abstract: Transformers have revolutionized machine learning across diverse domains, yet understanding their behavior remains crucial, particularly in high-stakes applications. This paper introduces the contextual counting task, a novel toy problem aimed at enhancing our understanding of Transformers in quantitative and scientific contexts. This task requires precise localization and computation within datasets, akin to object detection or region-based scientific analysis. We present theoretical and empirical analysis using both causal and non-causal Transformer architectures, investigating the influence of various positional encodings on performance and interpretability. In particular, we find that causal attention is much better suited for the task, and that no positional embeddings lead to the best accuracy, though rotary embeddings are competitive and easier to train. We also show that out of distribution performance is tightly linked to which tokens it uses as a bias term.
Authors: Kiri Daust, Adam Monahan
Abstract: Adapting to the changing climate requires accurate local climate information, a computationally challenging problem. Recent studies have used Generative Adversarial Networks (GANs), a type of deep learning, to learn complex distributions and downscale climate variables efficiently. Capturing variability while downscaling is crucial for estimating uncertainty and characterising extreme events - critical information for climate adaptation. Since downscaling is an undetermined problem, many fine-scale states are physically consistent with the coarse-resolution state. To quantify this ill-posed problem, downscaling techniques should be stochastic, able to sample realisations from a high-resolution distribution conditioned on low-resolution input. Previous stochastic downscaling attempts have found substantial underdispersion, with models failing to represent the full distribution. We propose approaches to improve the stochastic calibration of GANs in three ways: a) injecting noise inside the network, b) adjusting the training process to explicitly account for the stochasticity, and c) using a probabilistic loss metric. We tested our models first on a synthetic dataset with known distributional properties, and then on a realistic downscaling scenario, predicting high-resolution wind components from low-resolution climate covariates. Injecting noise, on its own, substantially improved the quality of conditional and full distributions in tests with synthetic data, but performed less well for wind field downscaling, where models remained underdispersed. For wind downscaling, we found that adjusting the training method and including the probabilistic loss improved calibration. The best model, with all three changes, showed much improved skill at capturing the full variability of the high-resolution distribution and thus at characterising extremes.
Authors: Ivan Dubrovsky, Andrei Dmitrenko, Aleksei Dmitrenko, Nikita Serov, Vladimir Vinogradov
Abstract: Creation of nanomaterials with specific morphology remains a complex experimental process, even though there is a growing demand for these materials in various industry sectors. This study explores the potential of AI to predict the morphology of nanoparticles within the data availability constraints. For that, we first generated a new multi-modal dataset that is double the size of analogous studies. Then, we systematically evaluated performance of classical machine learning and large language models in prediction of nanomaterial shapes and sizes. Finally, we prototyped a text-to-image system, discussed the obtained empirical results, as well as the limitations and promises of existing approaches.
Authors: Jay Desai, Xiaobo Guo, Srinivasan H. Sengamedu
Abstract: The performance of Large Language Models has achieved superhuman breadth with unprecedented depth. At the same time, the language models are mostly black box models and the underlying mechanisms for performance have been evaluated using synthetic or mechanistic schemes. We extend current mechanistic schemes to incorporate Logic, memory, and nuances of Language such as latent structure. The proposed framework is called LOLAMEME and we provide two instantiations of LOLAMEME: LoLa and MeMe languages. We then consider two generative language model architectures: transformer-based GPT-2 and convolution-based Hyena. We propose the hybrid architecture T HEX and use LOLAMEME framework is used to compare three architectures. T HEX outperforms GPT-2 and Hyena on select tasks.
Authors: Xuexiong Luo, Jia Wu, Jian Yang, Shan Xue, Amin Beheshti, Quan Z. Sheng, David McAlpine, Paul Sowman, Alexis Giral, Philip S. Yu
Abstract: Exploring the complex structure of the human brain is crucial for understanding its functionality and diagnosing brain disorders. Thanks to advancements in neuroimaging technology, a novel approach has emerged that involves modeling the human brain as a graph-structured pattern, with different brain regions represented as nodes and the functional relationships among these regions as edges. Moreover, graph neural networks (GNNs) have demonstrated a significant advantage in mining graph-structured data. Developing GNNs to learn brain graph representations for brain disorder analysis has recently gained increasing attention. However, there is a lack of systematic survey work summarizing current research methods in this domain. In this paper, we aim to bridge this gap by reviewing brain graph learning works that utilize GNNs. We first introduce the process of brain graph modeling based on common neuroimaging data. Subsequently, we systematically categorize current works based on the type of brain graph generated and the targeted research problems. To make this research accessible to a broader range of interested researchers, we provide an overview of representative methods and commonly used datasets, along with their implementation sources. Finally, we present our insights on future research directions. The repository of this survey is available at \url{https://github.com/XuexiongLuoMQ/Awesome-Brain-Graph-Learning-with-GNNs}.
URLs: https://github.com/XuexiongLuoMQ/Awesome-Brain-Graph-Learning-with-GNNs
Authors: Hojoon Lee, Hyeonseo Cho, Hyunseung Kim, Donghu Kim, Dugki Min, Jaegul Choo, Clare Lyle
Abstract: This study investigates the loss of generalization ability in neural networks, revisiting warm-starting experiments from Ash & Adams. Our empirical analysis reveals that common methods designed to enhance plasticity by maintaining trainability provide limited benefits to generalization. While reinitializing the network can be effective, it also risks losing valuable prior knowledge. To this end, we introduce the Hare & Tortoise, inspired by the brain's complementary learning system. Hare & Tortoise consists of two components: the Hare network, which rapidly adapts to new information analogously to the hippocampus, and the Tortoise network, which gradually integrates knowledge akin to the neocortex. By periodically reinitializing the Hare network to the Tortoise's weights, our method preserves plasticity while retaining general knowledge. Hare & Tortoise can effectively maintain the network's ability to generalize, which improves advanced reinforcement learning algorithms on the Atari-100k benchmark. The code is available at https://github.com/dojeon-ai/hare-tortoise.
Authors: Karn Tiwari, N M Anoop Krishnan, A P Prathosh
Abstract: Neural operators extend data-driven models to map between infinite-dimensional functional spaces. While these operators perform effectively in either the time or frequency domain, their performance may be limited when applied to non-stationary spatial or temporal signals whose frequency characteristics change with time. Here, we introduce Complex Neural Operator (CoNO) that parameterizes the integral kernel using Fractional Fourier Transform (FrFT), better representing non-stationary signals in a complex-valued domain. Theoretically, we prove the universal approximation capability of CoNO. We perform an extensive empirical evaluation of CoNO on seven challenging partial differential equations (PDEs), including regular grids, structured meshes, and point clouds. Empirically, CoNO consistently attains state-of-the-art performance, showcasing an average relative gain of 10.9%. Further, CoNO exhibits superior performance, outperforming all other models in additional tasks such as zero-shot super-resolution and robustness to noise. CoNO also exhibits the ability to learn from small amounts of data -- giving the same performance as the next best model with just 60% of the training data. Altogether, CoNO presents a robust and superior model for modeling continuous dynamical systems, providing a fillip to scientific machine learning.
Authors: Vedant Khandelwal, Amit Sheth, Forest Agostinelli
Abstract: Pathfinding problems are found throughout robotics, computational science, and natural sciences. Traditional methods to solve these require training deep neural networks (DNNs) for each new problem domain, consuming substantial time and resources. This study introduces a novel foundation model, leveraging deep reinforcement learning to train heuristic functions that seamlessly adapt to new domains without further fine-tuning. Building upon DeepCubeA, we enhance the model by providing the heuristic function with the domain's state transition information, improving its adaptability. Utilizing a puzzle generator for the 15-puzzle action space variation domains, we demonstrate our model's ability to generalize and solve unseen domains. We achieve a strong correlation between learned and ground truth heuristic values across various domains, as evidenced by robust R-squared and Concordance Correlation Coefficient metrics. These results underscore the potential of foundation models to establish new standards in efficiency and adaptability for AI-driven solutions in complex pathfinding problems.
Authors: Mohammed Djameleddine Belgoumri, Mohamed Reda Bouadjenek, Sunil Aryal, Hakim Hacid
Abstract: Data-driven Artificial Intelligence (AI) systems trained using Machine Learning (ML) are shaping an ever-increasing (in size and importance) portion of our lives, including, but not limited to, recommendation systems, autonomous driving technologies, healthcare diagnostics, financial services, and personalized marketing. On the one hand, the outsized influence of these systems imposes a high standard of quality, particularly in the data used to train them. On the other hand, establishing and maintaining standards of Data Quality (DQ) becomes more challenging due to the proliferation of Edge Computing and Internet of Things devices, along with their increasing adoption for training and deploying ML models. The nature of the edge environment -- characterized by limited resources, decentralized data storage, and processing -- exacerbates data-related issues, making them more frequent, severe, and difficult to detect and mitigate. From these observations, it follows that DQ research for edge ML is a critical and urgent exploration track for the safety and robust usefulness of present and future AI systems. Despite this fact, DQ research for edge ML is still in its infancy. The literature on this subject remains fragmented and scattered across different research communities, with no comprehensive survey to date. Hence, this paper aims to fill this gap by providing a global view of the existing literature from multiple disciplines that can be grouped under the umbrella of DQ for edge ML. Specifically, we present a tentative definition of data quality in Edge computing, which we use to establish a set of DQ dimensions. We explore each dimension in detail, including existing solutions for mitigation.
Authors: David Restrepo, Chenwei Wu, Sebasti\'an Andr\'es Cajas, Luis Filipe Nakayama, Leo Anthony Celi, Diego M L\'opez
Abstract: Large-scale multi-modal deep learning models have revolutionized domains such as healthcare, highlighting the importance of computational power. However, in resource-constrained regions like Low and Middle-Income Countries (LMICs), limited access to GPUs and data poses significant challenges, often leaving CPUs as the sole resource. To address this, we advocate for leveraging vector embeddings to enable flexible and efficient computational methodologies, democratizing multimodal deep learning across diverse contexts. Our paper investigates the efficiency and effectiveness of using vector embeddings from single-modal foundation models and multi-modal Vision-Language Models (VLMs) for multimodal deep learning in low-resource environments, particularly in healthcare. Additionally, we propose a simple yet effective inference-time method to enhance performance by aligning image-text embeddings. Comparing these approaches with traditional methods, we assess their impact on computational efficiency and model performance using metrics like accuracy, F1-score, inference time, training time, and memory usage across three medical modalities: BRSET (ophthalmology), HAM10000 (dermatology), and SatelliteBench (public health). Our findings show that embeddings reduce computational demands without compromising model performance. Furthermore, our alignment method improves performance in medical tasks. This research promotes sustainable AI practices by optimizing resources in constrained environments, highlighting the potential of embedding-based approaches for efficient multimodal learning. Vector embeddings democratize multimodal deep learning in LMICs, particularly in healthcare, enhancing AI adaptability in varied use cases.
Authors: Weiguo Chen, Changjian Wang, Kele Xu, Yuan Yuan, Yanru Bai, Dongsong Zhang
Abstract: Cognitive Language Processing (CLP), situated at the intersection of Natural Language Processing (NLP) and cognitive science, plays a progressively pivotal role in the domains of artificial intelligence, cognitive intelligence, and brain science. Among the essential areas of investigation in CLP, Cognitive Signal Decoding (CSD) has made remarkable achievements, yet there still exist challenges related to insufficient global dynamic representation capability and deficiencies in multi-domain feature integration. In this paper, we introduce a novel paradigm for CLP referred to as Disentangled Frequency-Spatial-Temporal Attention(D-FaST). Specifically, we present an novel cognitive signal decoder that operates on disentangled frequency-space-time domain attention. This decoder encompasses three key components: frequency domain feature extraction employing multi-view attention, spatial domain feature extraction utilizing dynamic brain connection graph attention, and temporal feature extraction relying on local time sliding window attention. These components are integrated within a novel disentangled framework. Additionally, to encourage advancements in this field, we have created a new CLP dataset, MNRED. Subsequently, we conducted an extensive series of experiments, evaluating D-FaST's performance on MNRED, as well as on publicly available datasets including ZuCo, BCIC IV-2A, and BCIC IV-2B. Our experimental results demonstrate that D-FaST outperforms existing methods significantly on both our datasets and traditional CSD datasets including establishing a state-of-the-art accuracy score 78.72% on MNRED, pushing the accuracy score on ZuCo to 78.35%, accuracy score on BCIC IV-2A to 74.85% and accuracy score on BCIC IV-2B to 76.81%.
Authors: Bivas Dinda
Abstract: The recent advancement of deep learning architectures, neural networks, and the combination of abundant financial data and powerful computers are transforming finance, leading us to develop an advanced method for predicting future stock prices. However, the accessibility of investment and trading at everyone's fingertips made the stock markets increasingly intricate and prone to volatility. The increased complexity and volatility of the stock market have driven demand for more models, which would effectively capture high volatility and non-linear behavior of the different stock prices. This study explored gated recurrent neural network (GRNN) algorithms such as LSTM (long short-term memory), GRU (gated recurrent unit), and hybrid models like GRU-LSTM, LSTM-GRU, with Tree-structured Parzen Estimator (TPE) Bayesian optimization for hyperparameter optimization (TPE-GRNN). The aim is to improve the prediction accuracy of the next day's closing price of the NIFTY 50 index, a prominent Indian stock market index, using TPE-GRNN. A combination of eight influential factors is carefully chosen from fundamental stock data, technical indicators, crude oil price, and macroeconomic data to train the models for capturing the changes in the price of the index with the factors of the broader economy. Single-layer and multi-layer TPE-GRNN models have been developed. The models' performance is evaluated using standard matrices like R2, MAPE, and RMSE. The analysis of models' performance reveals the impact of feature selection and hyperparameter optimization (HPO) in enhancing stock index price prediction accuracy. The results show that the MAPE of our proposed TPE-LSTM method is the lowest (best) with respect to all the previous models for stock index price prediction.
Authors: Jiayao Tan, Fan Lyu, Chenggong Ni, Tingliang Feng, Fuyuan Hu, Zhang Zhang, Shaochuang Zhao, Liang Wang
Abstract: Continual Test-Time Adaptation (CTTA) aims to adapt a pre-trained model to a sequence of target domains during the test phase without accessing the source data. To adapt to unlabeled data from unknown domains, existing methods rely on constructing pseudo-labels for all samples and updating the model through self-training. However, these pseudo-labels often involve noise, leading to insufficient adaptation. To improve the quality of pseudo-labels, we propose a pseudo-label selection method for CTTA, called Pseudo Labeling Filter (PLF). The key idea of PLF is to keep selecting appropriate thresholds for pseudo-labels and identify reliable ones for self-training. Specifically, we present three principles for setting thresholds during continuous domain learning, including initialization, growth and diversity. Based on these principles, we design Self-Adaptive Thresholding to filter pseudo-labels. Additionally, we introduce a Class Prior Alignment (CPA) method to encourage the model to make diverse predictions for unknown domain samples. Through extensive experiments, PLF outperforms current state-of-the-art methods, proving its effectiveness in CTTA.
Authors: Zikun Ye, Hema Yoganarasimhan, Yufeng Zheng
Abstract: In the rapidly evolving digital content landscape, media firms and news publishers require automated and efficient methods to enhance user engagement. This paper introduces the LLM-Assisted Online Learning Algorithm (LOLA), a novel framework that integrates Large Language Models (LLMs) with adaptive experimentation to optimize content delivery. Leveraging a large-scale dataset from Upworthy, which includes 17,681 headline A/B tests aimed at evaluating the performance of various headlines associated with the same article content, we first investigate three broad pure-LLM approaches: prompt-based methods, embedding-based classification models, and fine-tuned open-source LLMs. Our findings indicate that prompt-based approaches perform poorly, achieving no more than 65% accuracy in identifying the catchier headline among two options. In contrast, OpenAI-embedding-based classification models and fine-tuned Llama-3-8b models achieve comparable accuracy, around 82-84%, though still falling short of the performance of experimentation with sufficient traffic. We then introduce LOLA, which combines the best pure-LLM approach with the Upper Confidence Bound algorithm to adaptively allocate traffic and maximize clicks. Our numerical experiments on Upworthy data show that LOLA outperforms the standard A/B testing method (the current status quo at Upworthy), pure bandit algorithms, and pure-LLM approaches, particularly in scenarios with limited experimental traffic or numerous arms. Our approach is both scalable and broadly applicable to content experiments across a variety of digital settings where firms seek to optimize user engagement, including digital advertising and social media recommendations.
Authors: Ou Wu, Weiyao Zhu, Mengyang Li
Abstract: Measuring the value of individual samples is critical for many data-driven tasks, e.g., the training of a deep learning model. Recent literature witnesses the substantial efforts in developing data valuation methods. The primary data valuation methodology is based on the Shapley value from game theory, and various methods are proposed along this path. {Even though Shapley value-based valuation has solid theoretical basis, it is entirely an experiment-based approach and no valuation model has been constructed so far.} In addition, current data valuation methods ignore the interpretability of the output values, despite an interptable data valuation method is of great helpful for applications such as data pricing. This study aims to answer an important question: is data valuation learnable and interpretable? A learned valuation model have several desirable merits such as fixed number of parameters and knowledge reusability. An intrepretable data valuation model can explain why a sample is valuable or invaluable. To this end, two new data value modeling frameworks are proposed, in which a multi-layer perception~(MLP) and a new regression tree are utilized as specific base models for model training and interpretability, respectively. Extensive experiments are conducted on benchmark datasets. {The experimental results provide a positive answer for the question.} Our study opens up a new technical path for the assessing of data values. Large data valuation models can be built across many different data-driven tasks, which can promote the widespread application of data valuation.
Authors: Adel Nabli (MLIA, Mila), Louis Fournier (MLIA), Pierre Erbacher (MLIA), Louis Serrano (MLIA), Eugene Belilovsky (Mila), Edouard Oyallon
Abstract: Training Large Language Models (LLMs) relies heavily on distributed implementations, employing multiple GPUs to compute stochastic gradients on model replicas in parallel. However, synchronizing gradients in data parallel settings induces a communication overhead increasing with the number of distributed workers, which can impede the efficiency gains of parallelization. To address this challenge, optimization algorithms reducing inter-worker communication have emerged, such as local optimization methods used in Federated Learning. While effective in minimizing communication overhead, these methods incur significant memory costs, hindering scalability: in addition to extra momentum variables, if communications are only allowed between multiple local optimization steps, then the optimizer's states cannot be sharded among workers. In response, we propose $\textbf{AC}$cumulate while $\textbf{CO}$mmunicate ($\texttt{ACCO}$), a memory-efficient optimization algorithm tailored for distributed training of LLMs. $\texttt{ACCO}$ allows to shard optimizer states across workers, overlaps gradient computations and communications to conceal communication costs, and accommodates heterogeneous hardware. Our method relies on a novel technique to mitigate the one-step delay inherent in parallel execution of gradient computations and communications, eliminating the need for warmup steps and aligning with the training dynamics of standard distributed optimization while converging faster in terms of wall-clock time. We demonstrate the effectiveness of $\texttt{ACCO}$ on several LLMs training and fine-tuning tasks.
Authors: Zhanyu Liu, Jianrong Ding, Guanjie Zheng
Abstract: The field of Intelligent Transportation Systems (ITS) relies on accurate traffic forecasting to enable various downstream applications. However, developing cities often face challenges in collecting sufficient training traffic data due to limited resources and outdated infrastructure. Recognizing this obstacle, the concept of cross-city few-shot forecasting has emerged as a viable approach. While previous cross-city few-shot forecasting methods ignore the frequency similarity between cities, we have made an observation that the traffic data is more similar in the frequency domain between cities. Based on this fact, we propose a \textbf{F}requency \textbf{E}nhanced \textbf{P}re-training Framework for \textbf{Cross}-city Few-shot Forecasting (\textbf{FEPCross}). FEPCross has a pre-training stage and a fine-tuning stage. In the pre-training stage, we propose a novel Cross-Domain Spatial-Temporal Encoder that incorporates the information of the time and frequency domain and trains it with self-supervised tasks encompassing reconstruction and contrastive objectives. In the fine-tuning stage, we design modules to enrich training samples and maintain a momentum-updated graph structure, thereby mitigating the risk of overfitting to the few-shot training data. Empirical evaluations performed on real-world traffic datasets validate the exceptional efficacy of FEPCross, outperforming existing approaches of diverse categories and demonstrating characteristics that foster the progress of cross-city few-shot forecasting.
Authors: Victor Matray (LMPS), Faisal Amlani (LMPS), Fr\'ed\'eric Feyel (LMPS), David N\'eron (LMPS)
Abstract: This work introduces a new approach for accelerating the numerical analysis of time-domain partial differential equations (PDEs) governing complex physical systems. The methodology is based on a combination of a classical reduced-order modeling (ROM) framework and recently-introduced Graph Neural Networks (GNNs), where the latter is trained on highly heterogeneous databases of varying numerical discretization sizes. The proposed techniques are shown to be particularly suitable for non-parametric geometries, ultimately enabling the treatment of a diverse range of geometries and topologies. Performance studies are presented in an application context related to the design of aircraft seats and their corresponding mechanical responses to shocks, where the main motivation is to reduce the computational burden and enable the rapid design iteration for such problems that entail non-parametric geometries. The methods proposed here are straightforwardly applicable to other scientific or engineering problems requiring a large number of finite element-based numerical simulations, with the potential to significantly enhance efficiency while maintaining reasonable accuracy.
Authors: Yuxuan Chen, Rongpeng Li, Xiaoxue Yu, Zhifeng Zhao, Honggang Zhang
Abstract: Optimizing the deployment of large language models (LLMs) in edge computing environments is critical for enhancing privacy and computational efficiency. Toward efficient wireless LLM inference in edge computing, this study comprehensively analyzes the impact of different splitting points in mainstream open-source LLMs. On this basis, this study introduces a framework taking inspiration from model-based reinforcement learning (MBRL) to determine the optimal splitting point across the edge and user equipment (UE). By incorporating a reward surrogate model, our approach significantly reduces the computational cost of frequent performance evaluations. Extensive simulations demonstrate that this method effectively balances inference performance and computational load under varying network conditions, providing a robust solution for LLM deployment in decentralized settings.
Authors: Sanjay Kariyappa, Freddy L\'ecu\'e, Saumitra Mishra, Christopher Pond, Daniele Magazzeni, Manuela Veloso
Abstract: This paper proposes Progressive Inference - a framework to compute input attributions to explain the predictions of decoder-only sequence classification models. Our work is based on the insight that the classification head of a decoder-only Transformer model can be used to make intermediate predictions by evaluating them at different points in the input sequence. Due to the causal attention mechanism, these intermediate predictions only depend on the tokens seen before the inference point, allowing us to obtain the model's prediction on a masked input sub-sequence, with negligible computational overheads. We develop two methods to provide sub-sequence level attributions using this insight. First, we propose Single Pass-Progressive Inference (SP-PI), which computes attributions by taking the difference between consecutive intermediate predictions. Second, we exploit a connection with Kernel SHAP to develop Multi Pass-Progressive Inference (MP-PI). MP-PI uses intermediate predictions from multiple masked versions of the input to compute higher quality attributions. Our studies on a diverse set of models trained on text classification tasks show that SP-PI and MP-PI provide significantly better attributions compared to prior work.
Authors: Peiliang Gong, Mohamed Ragab, Emadeldeen Eldele, Wenyu Zhang, Min Wu, Chuan-Sheng Foo, Daoqiang Zhang, Xiaoli Li, Zhenghua Chen
Abstract: Source-free domain adaptation (SFDA) aims to adapt a model pre-trained on a labeled source domain to an unlabeled target domain without access to source data, preserving the source domain's privacy. While SFDA is prevalent in computer vision, it remains largely unexplored in time series analysis. Existing SFDA methods, designed for visual data, struggle to capture the inherent temporal dynamics of time series, hindering adaptation performance. This paper proposes MAsk And imPUte (MAPU), a novel and effective approach for time series SFDA. MAPU addresses the critical challenge of temporal consistency by introducing a novel temporal imputation task. This task involves randomly masking time series signals and leveraging a dedicated temporal imputer to recover the original signal within the learned embedding space, bypassing the complexities of noisy raw data. Notably, MAPU is the first method to explicitly address temporal consistency in the context of time series SFDA. Additionally, it offers seamless integration with existing SFDA methods, providing greater flexibility. We further introduce E-MAPU, which incorporates evidential uncertainty estimation to address the overconfidence issue inherent in softmax predictions. To achieve that, we leverage evidential deep learning to obtain a better-calibrated pre-trained model and adapt the target encoder to map out-of-support target samples to a new feature representation closer to the source domain's support. This fosters better alignment, ultimately enhancing adaptation performance. Extensive experiments on five real-world time series datasets demonstrate that both MAPU and E-MAPU achieve significant performance gains compared to existing methods. These results highlight the effectiveness of our proposed approaches for tackling various time series domain adaptation problems.
Authors: Yuda Wang, Xuxin He, Shengxin Zhu
Abstract: Sequential recommendation aims to estimate dynamic user preferences and sequential dependencies among historical user behaviors. Attention-based models have proven effective for sequential recommendation, but they suffer from inference inefficiency due to the quadratic computational complexity of attention mechanisms, particularly for long-range behavior sequences. Inspired by the recent success of state space models (SSMs) in control theory, which provide a robust framework for modeling and controlling dynamic systems, we present EchoMamba4Rec. Control theory emphasizes the use of SSMs for managing long-range dependencies and maintaining inferential efficiency through structured state matrices. EchoMamba4Rec leverages these control relationships in sequential recommendation and integrates bi-directional processing with frequency-domain filtering to capture complex patterns and dependencies in user interaction data more effectively. Our model benefits from the ability of state space models (SSMs) to learn and perform parallel computations, significantly enhancing computational efficiency and scalability. It features a bi-directional Mamba module that incorporates both forward and reverse Mamba components, leveraging information from both past and future interactions. Additionally, a filter layer operates in the frequency domain using learnable Fast Fourier Transform (FFT) and learnable filters, followed by an inverse FFT to refine item embeddings and reduce noise. We also integrate Gate Linear Units (GLU) to dynamically control information flow, enhancing the model's expressiveness and training stability. Experimental results demonstrate that EchoMamba significantly outperforms existing models, providing more accurate and personalized recommendations.
Authors: Zhou Yang, Zhaochun Ren, Chenglong Ye, Yufeng Wang, Haizhou Sun, Chao Chen, Xiaofei Zhu, Yunbing Wu, Xiangwen Liao
Abstract: In-context learning (ICL) achieves remarkable performance in various domains such as knowledge acquisition, commonsense reasoning, and semantic understanding. However, its performance significantly deteriorates for emotion detection tasks, especially fine-grained emotion recognition. The underlying reasons for this remain unclear. In this paper, we identify the reasons behind ICL's poor performance from the perspective of prototype theory and propose a method to address this issue. Specifically, we conduct extensive pilot experiments and find that ICL conforms to the prototype theory on fine-grained emotion recognition. Based on this theory, we uncover the following deficiencies in ICL: (1) It relies on prototypes (example-label pairs) that are semantically similar but emotionally inaccurate to predict emotions. (2) It is prone to interference from irrelevant categories, affecting the accuracy and robustness of the predictions. To address these issues, we propose an Emotion Context Learning method (E-ICL) on fine-grained emotion recognition. E-ICL relies on more emotionally accurate prototypes to predict categories by referring to emotionally similar examples with dynamic labels. Simultaneously, E-ICL employs an exclusionary emotion prediction strategy to avoid interference from irrelevant categories, thereby increasing its accuracy and robustness. Note that the entire process is accomplished with the assistance of a plug-and-play emotion auxiliary model, without additional training. Experiments on the fine-grained emotion datasets EDOS, Empathetic-Dialogues, EmpatheticIntent, and GoEmotions show that E-ICL achieves superior emotion prediction performance. Furthermore, even when the emotion auxiliary model used is lower than 10% of the LLMs, E-ICL can still boost the performance of LLMs by over 4% on multiple datasets.
Authors: Vojtech Halenka, Ahmed K. Kadhim, Paul F. A. Clarke, Bimal Bhattarai, Rupsa Saha, Ole-Christoffer Granmo, Lei Jiao, Per-Arne Andersen
Abstract: Tsetlin machines (TMs) have been successful in several application domains, operating with high efficiency on Boolean representations of the input data. However, Booleanizing complex data structures such as sequences, graphs, images, signal spectra, chemical compounds, and natural language is not trivial. In this paper, we propose a hypervector (HV) based method for expressing arbitrarily large sets of concepts associated with any input data. Using a hyperdimensional space to build vectors drastically expands the capacity and flexibility of the TM. We demonstrate how images, chemical compounds, and natural language text are encoded according to the proposed method, and how the resulting HV-powered TM can achieve significantly higher accuracy and faster learning on well-known benchmarks. Our results open up a new research direction for TMs, namely how to expand and exploit the benefits of operating in hyperspace, including new booleanization strategies, optimization of TM inference and learning, as well as new TM applications.
Authors: Michael Schlechtinger, Damaris Kosack, Franz Krause, Heiko Paulheim
Abstract: In the rapidly evolving landscape of eCommerce, Artificial Intelligence (AI) based pricing algorithms, particularly those utilizing Reinforcement Learning (RL), are becoming increasingly prevalent. This rise has led to an inextricable pricing situation with the potential for market collusion. Our research employs an experimental oligopoly model of repeated price competition, systematically varying the environment to cover scenarios from basic economic theory to subjective consumer demand preferences. We also introduce a novel demand framework that enables the implementation of various demand models, allowing for a weighted blending of different models. In contrast to existing research in this domain, we aim to investigate the strategies and emerging pricing patterns developed by the agents, which may lead to a collusive outcome. Furthermore, we investigate a scenario where agents cannot observe their competitors' prices. Finally, we provide a comprehensive legal analysis across all scenarios. Our findings indicate that RL-based AI agents converge to a collusive state characterized by the charging of supracompetitive prices, without necessarily requiring inter-agent communication. Implementing alternative RL algorithms, altering the number of agents or simulation settings, and restricting the scope of the agents' observation space does not significantly impact the collusive market outcome behavior.
Authors: Yunbo Hou, Haoran Ye, Yingxue Zhang, Siyuan Xu, Guojie Song
Abstract: Placement is a critical and challenging step of modern chip design, with routability being an essential indicator of placement quality. Current routability-oriented placers typically apply an iterative two-stage approach, wherein the first stage generates a placement solution, and the second stage provides non-differentiable routing results to heuristically improve the solution quality. This method hinders jointly optimizing the routability aspect during placement. To address this problem, this work introduces RoutePlacer, an end-to-end routability-aware placement method. It trains RouteGNN, a customized graph neural network, to efficiently and accurately predict routability by capturing and fusing geometric and topological representations of placements. Well-trained RouteGNN then serves as a differentiable approximation of routability, enabling end-to-end gradient-based routability optimization. In addition, RouteGNN can improve two-stage placers as a plug-and-play alternative to external routers. Our experiments on DREAMPlace, an open-source AI4EDA platform, show that RoutePlacer can reduce Total Overflow by up to 16% while maintaining routed wirelength, compared to the state-of-the-art; integrating RouteGNN within two-stage placers leads to a 44% reduction in Total Overflow without compromising wirelength.
Authors: Aidan Scannell, Kalle Kujanp\"a\"a, Yi Zhao, Mohammadreza Nakhaei, Arno Solin, Joni Pajarinen
Abstract: Learning representations for reinforcement learning (RL) has shown much promise for continuous control. We propose an efficient representation learning method using only a self-supervised latent-state consistency loss. Our approach employs an encoder and a dynamics model to map observations to latent states and predict future latent states, respectively. We achieve high performance and prevent representation collapse by quantizing the latent representation such that the rank of the representation is empirically preserved. Our method, named iQRL: implicitly Quantized Reinforcement Learning, is straightforward, compatible with any model-free RL algorithm, and demonstrates excellent performance by outperforming other recently proposed representation learning methods in continuous control benchmarks from DeepMind Control Suite.
Authors: Scott H. Hawley, Austin R. Tackett
Abstract: We investigate the construction of latent spaces through self-supervised learning to support semantically meaningful operations. Analogous to operational amplifiers, these "operational latent spaces" (OpLaS) not only demonstrate semantic structure such as clustering but also support common transformational operations with inherent semantic meaning. Some operational latent spaces are found to have arisen "unintentionally" in the progress toward some (other) self-supervised learning objective, in which unintended but still useful properties are discovered among the relationships of points in the space. Other spaces may be constructed "intentionally" by developers stipulating certain kinds of clustering or transformations intended to produce the desired structure. We focus on the intentional creation of operational latent spaces via self-supervised learning, including the introduction of rotation operators via a novel "FiLMR" layer, which can be used to enable ring-like symmetries found in some musical constructions.
Authors: Aram Avetisyan, Nikolas Khachaturov, Ariana Asatryan, Shahane Tigranyan, Yury Markin
Abstract: Electrocardiogram (ECG) delineation plays a crucial role in assisting cardiologists with accurate diagnoses. Prior research studies have explored various methods, including the application of deep learning techniques, to achieve precise delineation. However, existing approaches face limitations primarily related to dataset size and robustness. In this paper, we introduce a dataset for ECG delineation and propose a novel self-trained method aimed at leveraging a vast amount of unlabeled ECG data. Our approach involves the pseudolabeling of unlabeled data using a neural network trained on our dataset. Subsequently, we train the model on the newly labeled samples to enhance the quality of delineation. We conduct experiments demonstrating that our dataset is a valuable resource for training robust models and that our proposed self-trained method improves the prediction quality of ECG delineation.
Authors: Christopher A. Choquette-Choo, Arun Ganesh, Abhradeep Thakurta
Abstract: The most common algorithms for differentially private (DP) machine learning (ML) are all based on stochastic gradient descent, for example, DP-SGD. These algorithms achieve DP by treating each gradient as an independent private query. However, this independence can cause us to overpay in privacy loss because we don't analyze the entire gradient trajectory. In this work, we propose a new DP algorithm, which we call Accelerated-DP-SRGD (DP stochastic recursive gradient descent), that enables us to break this independence and only pay for privacy in the gradient difference, i.e., in the new information at the current step. Our algorithm achieves the optimal DP-stochastic convex optimization (DP-SCO) error (up to polylog factors) using only a single epoch over the dataset, and converges at the Nesterov's accelerated rate. Our algorithm can be run in at most $\sqrt{n}$ batch gradient steps with batch size at least $\sqrt{n}$, unlike prior work which required $O(n)$ queries with mostly constant batch sizes. To achieve this, our algorithm combines three key ingredients, a variant of stochastic recursive gradients (SRG), accelerated gradient descent, and correlated noise generation from DP continual counting. Finally, we also show that our algorithm improves over existing SoTA on multi-class logistic regression on MNIST and CIFAR-10.
Authors: Sanghyun Lee, Chanyoung Park
Abstract: Accurate traffic flow forecasting is a crucial research topic in transportation management. However, it is a challenging problem due to rapidly changing traffic conditions, high nonlinearity of traffic flow, and complex spatial and temporal correlations of road networks. Most existing studies either try to capture the spatial dependencies between roads using the same semantic graph over different time steps, or assume all sensors on the roads are equally likely to be connected regardless of the distance between them. However, we observe that the spatial dependencies between roads indeed change over time, and two distant roads are not likely to be helpful to each other when predicting the traffic flow, both of which limit the performance of existing studies. In this paper, we propose Temporal Graph Learning Recurrent Neural Network (TGLRN) to address these problems. More precisely, to effectively model the nature of time series, we leverage Recurrent Neural Networks (RNNs) to dynamically construct a graph at each time step, thereby capturing the time-evolving spatial dependencies between roads (i.e., microscopic view). Simultaneously, we provide the Adaptive Structure Information to the model, ensuring that close and consecutive sensors are considered to be more important for predicting the traffic flow (i.e., macroscopic view). Furthermore, to endow TGLRN with robustness, we introduce an edge sampling strategy when constructing the graph at each time step, which eventually leads to further improvements on the model performance. Experimental results on four commonly used real-world benchmark datasets show the effectiveness of TGLRN.
Authors: Simon Zhang, Soham Mukherjee, Tamal K. Dey
Abstract: Extended persistence is a technique from topological data analysis to obtain global multiscale topological information from a graph. This includes information about connected components and cycles that are captured by the so-called persistence barcodes. We introduce extended persistence into a supervised learning framework for graph classification. Global topological information, in the form of a barcode with four different types of bars and their explicit cycle representatives, is combined into the model by the readout function which is computed by extended persistence. The entire model is end-to-end differentiable. We use a link-cut tree data structure and parallelism to lower the complexity of computing extended persistence, obtaining a speedup of more than 60x over the state-of-the-art for extended persistence computation. This makes extended persistence feasible for machine learning. We show that, under certain conditions, extended persistence surpasses both the WL[1] graph isomorphism test and 0-dimensional barcodes in terms of expressivity because it adds more global (topological) information. In particular, arbitrarily long cycles can be represented, which is difficult for finite receptive field message passing graph neural networks. Furthermore, we show the effectiveness of our method on real world datasets compared to many existing recent graph representation learning methods.
Authors: Carolina Trindade, Lu\'is Antunes, T\^ania Carvalho, Nuno Moniz
Abstract: Multiple synthetic data generation models have emerged, among which deep learning models have become the vanguard due to their ability to capture the underlying characteristics of the original data. However, the resemblance of the synthetic to the original data raises important questions on the protection of individuals' privacy. As synthetic data is perceived as a means to fully protect personal information, most current related work disregards the impact of re-identification risk. In particular, limited attention has been given to exploring outliers, despite their privacy relevance. In this work, we analyze the privacy of synthetic data w.r.t the outliers. Our main findings suggest that outliers re-identification via linkage attack is feasible and easily achieved. Furthermore, additional safeguards such as differential privacy can prevent re-identification, albeit at the expense of the data utility.
Authors: Alessio Gravina, Giulio Lovisotto, Claudio Gallicchio, Davide Bacciu, Claas Grohnfeldt
Abstract: Learning Continuous-Time Dynamic Graphs (C-TDGs) requires accurately modeling spatio-temporal information on streams of irregularly sampled events. While many methods have been proposed recently, we find that most message passing-, recurrent- or self-attention-based methods perform poorly on long-range tasks. These tasks require correlating information that occurred "far" away from the current event, either spatially (higher-order node information) or along the time dimension (events occurred in the past). To address long-range dependencies, we introduce Continuous-Time Graph Anti-Symmetric Network (CTAN). Grounded within the ordinary differential equations framework, our method is designed for efficient propagation of information. In this paper, we show how CTAN's (i) long-range modeling capabilities are substantiated by theoretical findings and how (ii) its empirical performance on synthetic long-range benchmarks and real-world benchmarks is superior to other methods. Our results motivate CTAN's ability to propagate long-range information in C-TDGs as well as the inclusion of long-range tasks as part of temporal graph models evaluation.
Authors: Nathan Ng, Roger Grosse, Marzyeh Ghassemi
Abstract: Estimating the uncertainty of a model's prediction on a test point is a crucial part of ensuring reliability and calibration under distribution shifts. A minimum description length approach to this problem uses the predictive normalized maximum likelihood (pNML) distribution, which considers every possible label for a data point, and decreases confidence in a prediction if other labels are also consistent with the model and training data. In this work we propose IF-COMP, a scalable and efficient approximation of the pNML distribution that linearizes the model with a temperature-scaled Boltzmann influence function. IF-COMP can be used to produce well-calibrated predictions on test points as well as measure complexity in both labelled and unlabelled settings. We experimentally validate IF-COMP on uncertainty calibration, mislabel detection, and OOD detection tasks, where it consistently matches or beats strong baseline methods.
Authors: Ilgee Hong, Zichong Li, Alexander Bukharin, Yixiao Li, Haoming Jiang, Tianbao Yang, Tuo Zhao
Abstract: Reinforcement learning from human feedback (RLHF) is a prevalent approach to align AI systems with human values by learning rewards from human preference data. Due to various reasons, however, such data typically takes the form of rankings over pairs of trajectory segments, which fails to capture the varying strengths of preferences across different pairs. In this paper, we propose a novel adaptive preference loss, underpinned by distributionally robust optimization (DRO), designed to address this uncertainty in preference strength. By incorporating an adaptive scaling parameter into the loss for each pair, our method increases the flexibility of the reward function. Specifically, it assigns small scaling parameters to pairs with ambiguous preferences, leading to more comparable rewards, and large scaling parameters to those with clear preferences for more distinct rewards. Computationally, our proposed loss function is strictly convex and univariate with respect to each scaling parameter, enabling its efficient optimization through a simple second-order algorithm. Our method is versatile and can be readily adapted to various preference optimization frameworks, including direct preference optimization (DPO). Our experiments with robotic control and natural language generation with large language models (LLMs) show that our method not only improves policy performance but also aligns reward function selection more closely with policy optimization, simplifying the hyperparameter tuning process.
Authors: Kathrin Donandt, Dirk S\"offker
Abstract: Autonomous transportation systems such as road vehicles or vessels require the consideration of the static and dynamic environment to dislocate without collision. Anticipating the behavior of an agent in a given situation is required to adequately react to it in time. Developing deep learning-based models has become the dominant approach to motion prediction recently. The social environment is often considered through a CNN-LSTM-based sub-module processing a $\textit{social tensor}$ that includes information of the past trajectory of surrounding agents. For the proposed transformer-based trajectory prediction model, an alternative, computationally more efficient social tensor definition and processing is suggested. It considers the interdependencies between target and surrounding agents at each time step directly instead of relying on information of last hidden LSTM states of individually processed agents. A transformer-based sub-module, the Social Tensor Transformer, is integrated into the overall prediction model. It is responsible for enriching the target agent's dislocation features with social interaction information obtained from the social tensor. For the awareness of spatial limitations, dislocation features are defined in relation to the navigable area. This replaces additional, computationally expensive map processing sub-modules. An ablation study shows, that for longer prediction horizons, the deviation of the predicted trajectory from the ground truth is lower compared to a spatially and socially agnostic model. Even if the performance gain from a spatial-only to a spatial and social context-sensitive model is small in terms of common error measures, by visualizing the results it can be shown that the proposed model in fact is able to predict reactions to surrounding agents and explicitely allows an interpretable behavior.
Authors: Kathrin Donandt, Karim B\"ottger, Dirk S\"offker
Abstract: Accurate vessel trajectory prediction is necessary for save and efficient navigation. Deep learning-based prediction models, esp. encoder-decoders, are rarely applied to inland navigation specifically. Approaches from the maritime domain cannot directly be transferred to river navigation due to specific driving behavior influencing factors. Different encoder-decoder architectures, including a transformer encoder-decoder, are compared herein for predicting the next positions of inland vessels, given not only spatio-temporal information from AIS, but also river specific features. The results show that the reformulation of the regression task as classification problem and the inclusion of river specific features yield the lowest displacement errors. The standard LSTM encoder-decoder outperforms the transformer encoder-decoder for the data considered, but is computationally more expensive. In this study for the first time a transformer-based encoder-decoder model is applied to the problem of predicting the ship trajectory. Here, a feature vector using the river-specific context of navigation input parameters is established. Future studies can built on the proposed models, investigate the improvement of the computationally more efficient transformer, e.g. through further hyper-parameter optimization, and use additional river-specific information in the context representation to further increase prediction accuracy.
Authors: Kathrin Donandt, Karim B\"ottger, Dirk S\"offker
Abstract: Physics-related and model-based vessel trajectory prediction is highly accurate but requires specific knowledge of the vessel under consideration which is not always practical. Machine learning-based trajectory prediction models do not require expert knowledge, but rely on the implicit knowledge extracted from massive amounts of data. Several deep learning (DL) methods for vessel trajectory prediction have recently been suggested. The DL models developed typically only process information about the (dis)location of vessels defined with respect to a global reference system. In the context of inland navigation, this can be problematic, since without knowledge of the limited navigable space, irrealistic trajectories are likely to be determined. If spatial constraintes are introduced, e.g., by implementing an additional submodule to process map data, however, overall complexity increases. Instead of processing the vessel displacement information on the one hand and the spatial information on the other hand, the paper proposes the merging of both information. Here, fairway-related and navigation-related displacement information are used directly. In this way, the previously proposed context-sensitive Classification Transformer (CSCT) shows an improved spatial awareness. Additionally, the CSCT is adapted to assess the model uncertainty by enabling dropout during inference. This approach is trained on different inland waterways to analyze its generalizability. As the improved CSCT obtains lower prediction errors and enables to estimate the trustworthiness of each prediction, it is more suitable for safety-critical applications in inland navigation than previously developed models.
Authors: Roya Aliakbarisani, Robert Jankowski, M. \'Angeles Serrano, Mari\'an Bogu\~n\'a
Abstract: Graph Neural Networks (GNNs) have excelled in predicting graph properties in various applications ranging from identifying trends in social networks to drug discovery and malware detection. With the abundance of new architectures and increased complexity, GNNs are becoming highly specialized when tested on a few well-known datasets. However, how the performance of GNNs depends on the topological and features properties of graphs is still an open question. In this work, we introduce a comprehensive benchmarking framework for graph machine learning, focusing on the performance of GNNs across varied network structures. Utilizing the geometric soft configuration model in hyperbolic space, we generate synthetic networks with realistic topological properties and node feature vectors. This approach enables us to assess the impact of network properties, such as topology-feature correlation, degree distributions, local density of triangles (or clustering), and homophily, on the effectiveness of different GNN architectures. Our results highlight the dependency of model performance on the interplay between network structure and node features, providing insights for model selection in various scenarios. This study contributes to the field by offering a versatile tool for evaluating GNNs, thereby assisting in developing and selecting suitable models based on specific data characteristics.
Authors: Advait Gadhikar, Sree Harsha Nelaturu, Rebekka Burkholz
Abstract: The success of iterative pruning methods in achieving state-of-the-art sparse networks has largely been attributed to improved mask identification and an implicit regularization induced by pruning. We challenge this hypothesis and instead posit that their repeated cyclic training schedules enable improved optimization. To verify this, we show that pruning at initialization is significantly boosted by repeated cyclic training, even outperforming standard iterative pruning methods. The dominant mechanism how this is achieved, as we conjecture, can be attributed to a better exploration of the loss landscape leading to a lower training loss. However, at high sparsity, repeated cyclic training alone is not enough for competitive performance. A strong coupling between learnt parameter initialization and mask seems to be required. Standard methods obtain this coupling via expensive pruning-training iterations, starting from a dense network. To achieve this with sparse training instead, we propose SCULPT-ing, i.e., repeated cyclic training of any sparse mask followed by a single pruning step to couple the parameters and the mask, which is able to match the performance of state-of-the-art iterative pruning methods in the high sparsity regime at reduced computational cost.
Authors: Florian Stadtmann, Adil Rasheed
Abstract: The demand for condition-based and predictive maintenance is rising across industries, especially for remote, high-value, and high-risk assets. In this article, the diagnostic digital twin concept is introduced, discussed, and implemented for a floating offshore turbine. A diagnostic digital twin is a virtual representation of an asset that combines real-time data and models to monitor damage, detect anomalies, and diagnose failures, thereby enabling condition-based and predictive maintenance. By applying diagnostic digital twins to offshore assets, unexpected failures can be alleviated, but the implementation can prove challenging. Here, a diagnostic digital twin is implemented for an operational floating offshore wind turbine. The asset is monitored through measurements. Unsupervised learning methods are employed to build a normal operation model, detect anomalies, and provide a fault diagnosis. Warnings and diagnoses are sent through text messages, and a more detailed diagnosis can be accessed in a virtual reality interface. The diagnostic digital twin successfully detected an anomaly with high confidence hours before a failure occurred. The paper concludes by discussing diagnostic digital twins in the broader context of offshore engineering. The presented approach can be generalized to other offshore assets to improve maintenance and increase the lifetime, efficiency, and sustainability of offshore assets.
Authors: Shay Deutsch, Lionel Yelibi, Alex Tong Lin, Arjun Ravi Kannan
Abstract: Deriving meaningful representations from complex, high-dimensional data in unsupervised settings is crucial across diverse machine learning applications. This paper introduces a framework for multi-scale graph network embedding based on spectral graph wavelets that employs a contrastive learning approach. A significant feature of the proposed embedding is its capacity to establish a correspondence between the embedding space and the input feature space which aids in deriving feature importance of the original features. We theoretically justify our approach and demonstrate that, in Paley-Wiener spaces on combinatorial graphs, the spectral graph wavelets operator offers greater flexibility and better control over smoothness properties compared to the Laplacian operator. We validate the effectiveness of our proposed graph embedding on a variety of public datasets through a range of downstream tasks, including clustering and unsupervised feature importance.
Authors: Yejia Liu, Jianyi Yang, Pengfei Li, Tongxin Li, Shaolei Ren
Abstract: Public models offer predictions to a variety of downstream tasks and have played a crucial role in various AI applications, showcasing their proficiency in accurate predictions. However, the exclusive emphasis on prediction accuracy may not align with the diverse end objectives of downstream agents. Recognizing the public model's predictions as a service, we advocate for integrating the objectives of downstream agents into the optimization process. Concretely, to address performance disparities and foster fairness among heterogeneous agents in training, we propose a novel Equitable Objective. This objective, coupled with a policy gradient algorithm, is crafted to train the public model to produce a more equitable/uniform performance distribution across downstream agents, each with their unique concerns. Both theoretical analysis and empirical case studies have proven the effectiveness of our method in advancing performance equity across diverse downstream agents utilizing the public model for their decision-making. Codes and datasets are released at https://github.com/Ren-Research/Socially-Equitable-Public-Models.
URLs: https://github.com/Ren-Research/Socially-Equitable-Public-Models.
Authors: R\'obert Istv\'an Busa-Fekete, Travis Dick, Claudio Gentile, Andr\'es Mu\~noz Medina, Adam Smith, Marika Swanberg
Abstract: We propose reconstruction advantage measures to audit label privatization mechanisms. A reconstruction advantage measure quantifies the increase in an attacker's ability to infer the true label of an unlabeled example when provided with a private version of the labels in a dataset (e.g., aggregate of labels from different users or noisy labels output by randomized response), compared to an attacker that only observes the feature vectors, but may have prior knowledge of the correlation between features and labels. We consider two such auditing measures: one additive, and one multiplicative. These incorporate previous approaches taken in the literature on empirical auditing and differential privacy. The measures allow us to place a variety of proposed privatization schemes -- some differentially private, some not -- on the same footing. We analyze these measures theoretically under a distributional model which encapsulates reasonable adversarial settings. We also quantify their behavior empirically on real and simulated prediction tasks. Across a range of experimental settings, we find that differentially private schemes dominate or match the privacy-utility tradeoff of more heuristic approaches.
Authors: Yifei Wang, Sungyoon Kim, Paul Chu, Indu Subramaniam, Mert Pilanci
Abstract: We introduce randomized algorithms to Clifford's Geometric Algebra, generalizing randomized linear algebra to hypercomplex vector spaces. This novel approach has many implications in machine learning, including training neural networks to global optimality via convex optimization. Additionally, we consider fine-tuning large language model (LLM) embeddings as a key application area, exploring the intersection of geometric algebra and modern AI techniques. In particular, we conduct a comparative analysis of the robustness of transfer learning via embeddings, such as OpenAI GPT models and BERT, using traditional methods versus our novel approach based on convex optimization. We test our convex optimization transfer learning method across a variety of case studies, employing different embeddings (GPT-4 and BERT embeddings) and different text classification datasets (IMDb, Amazon Polarity Dataset, and GLUE) with a range of hyperparameter settings. Our results demonstrate that convex optimization and geometric algebra not only enhances the performance of LLMs but also offers a more stable and reliable method of transfer learning via embeddings.
Authors: Yuansan Liu, Sudanthi Wijewickrema, Dongting Hu, Christofer Bester, Stephen O'Leary, James Bailey
Abstract: Recent innovations in diffusion probabilistic models have paved the way for significant progress in image, text and audio generation, leading to their applications in generative time series forecasting. However, leveraging such abilities to model highly stochastic time series data remains a challenge. In this paper, we propose a novel Stochastic Diffusion (StochDiff) model which learns data-driven prior knowledge at each time step by utilizing the representational power of the stochastic latent spaces to model the variability of the multivariate time series data. The learnt prior knowledge helps the model to capture complex temporal dynamics and the inherent uncertainty of the data. This improves its ability to model highly stochastic time series data. Through extensive experiments on real-world datasets, we demonstrate the effectiveness of our proposed model on stochastic time series forecasting. Additionally, we showcase an application of our model for real-world surgical guidance, highlighting its potential to benefit the medical community.
Authors: Niccol\`o Ronchetti
Abstract: We propose Conditional Idempotent Generative Networks (CIGN), a novel approach that expands upon Idempotent Generative Networks (IGN) to enable conditional generation. While IGNs offer efficient single-pass generation, they lack the ability to control the content of the generated data. CIGNs address this limitation by incorporating conditioning mechanisms, allowing users to steer the generation process towards specific types of data. We establish the theoretical foundations for CIGNs, outlining their scope, loss function design, and evaluation metrics. We then present two potential architectures for implementing CIGNs: channel conditioning and filter conditioning. Finally, we discuss experimental results on the MNIST dataset, demonstrating the effectiveness of both approaches. Our findings pave the way for further exploration of CIGNs on larger datasets and with more powerful computing resources to determine the optimal implementation strategy.
Authors: Brian K Chen, Tianyang Hu, Hui Jin, Hwee Kuan Lee, Kenji Kawaguchi
Abstract: In-Context Learning (ICL) has been a powerful emergent property of large language models that has attracted increasing attention in recent years. In contrast to regular gradient-based learning, ICL is highly interpretable and does not require parameter updates. In this paper, we show that, for linearized transformer networks, ICL can be made explicit and permanent through the inclusion of bias terms. We mathematically demonstrate the equivalence between a model with ICL demonstration prompts and the same model with the additional bias terms. Our algorithm (ICLCA) allows for exact conversion in an inexpensive manner. Existing methods are not exact and require expensive parameter updates. We demonstrate the efficacy of our approach through experiments that show the exact incorporation of ICL tokens into a linear transformer. We further suggest how our method can be adapted to achieve cheap approximate conversion of ICL tokens, even in regular transformer networks that are not linearized. Our experiments on GPT-2 show that, even though the conversion is only approximate, the model still gains valuable context from the included bias terms.
Authors: Yuji Kawai, Takashi Morita, Jihoon Park, Minoru Asada
Abstract: Reservoir computing, a machine learning framework used for modeling the brain, can predict temporal data with little observations and minimal computational resources. However, it is difficult to accurately reproduce the long-term target time series because the reservoir system becomes unstable. This predictive capability is required for a wide variety of time-series processing, including predictions of motor timing and chaotic dynamical systems. This study proposes oscillation-driven reservoir computing (ODRC) with feedback, where oscillatory signals are fed into a reservoir network to stabilize the network activity and induce complex reservoir dynamics. The ODRC can reproduce long-term target time series more accurately than conventional reservoir computing methods in a motor timing and chaotic time-series prediction tasks. Furthermore, it generates a time series similar to the target in the unexperienced period, that is, it can learn the abstract generative rules from limited observations. Given these significant improvements made by the simple and computationally inexpensive implementation, the ODRC would serve as a practical model of various time series data. Moreover, we will discuss biological implications of the ODRC, considering it as a model of neural oscillations and their cerebellar processors.
Authors: Yang Liu, Peng Zhang, Yang Gao, Chuan Zhou, Zhao Li, Hongyang Chen
Abstract: In recent years, graph neural networks (GNNs) have become increasingly popular for solving NP-hard combinatorial optimization (CO) problems, such as maximum cut and maximum independent set. The core idea behind these methods is to represent a CO problem as a graph and then use GNNs to learn the node/graph embedding with combinatorial information. Although these methods have achieved promising results, given a specific CO problem, the design of GNN architectures still requires heavy manual work with domain knowledge. Existing automated GNNs are mostly focused on traditional graph learning problems, which is inapplicable to solving NP-hard CO problems. To this end, we present a new class of \textbf{AUTO}mated \textbf{G}NNs for solving \textbf{NP}-hard problems, namely \textbf{AutoGNP}. We represent CO problems by GNNs and focus on two specific problems, i.e., mixed integer linear programming and quadratic unconstrained binary optimization. The idea of AutoGNP is to use graph neural architecture search algorithms to automatically find the best GNNs for a given NP-hard combinatorial optimization problem. Compared with existing graph neural architecture search algorithms, AutoGNP utilizes two-hop operators in the architecture search space. Moreover, AutoGNP utilizes simulated annealing and a strict early stopping policy to avoid local optimal solutions. Empirical results on benchmark combinatorial problems demonstrate the superiority of our proposed model.
Authors: George Nehma, Madhur Tiwari
Abstract: Multi-layer perceptrons (MLP's) have been extensively utilized in discovering Deep Koopman operators for linearizing nonlinear dynamics. With the emergence of Kolmogorov-Arnold Networks (KANs) as a more efficient and accurate alternative to the MLP Neural Network, we propose a comparison of the performance of each network type in the context of learning Koopman operators with control.In this work, we propose a KANs-based deep Koopman framework with applications to an orbital Two-Body Problem (2BP) and the pendulum for data-driven discovery of linear system dynamics. KANs were found to be superior in nearly all aspects of training; learning 31 times faster, being 15 times more parameter efficiency, and predicting 1.25 times more accurately as compared to the MLP Deep Neural Networks (DNNs) in the case of the 2BP. Thus, KANs shows potential for being an efficient tool in the development of Deep Koopman Theory.
Authors: Jeffrey Ma, Alan Tu, Yiling Chen, Vijay Janapa Reddi
Abstract: Federated Learning (FL) endeavors to harness decentralized data while preserving privacy, facing challenges of performance, scalability, and collaboration. Asynchronous Federated Learning (AFL) methods have emerged as promising alternatives to their synchronous counterparts bounded by the slowest agent, yet they add additional challenges in convergence guarantees, fairness with respect to compute heterogeneity, and incorporation of staleness in aggregated updates. Specifically, AFL biases model training heavily towards agents who can produce updates faster, leaving slower agents behind, who often also have differently distributed data which is not learned by the global model. Naively upweighting introduces incentive issues, where true fast updating agents may falsely report updates at a slower speed to increase their contribution to model training. We introduce FedStaleWeight, an algorithm addressing fairness in aggregating asynchronous client updates by employing average staleness to compute fair re-weightings. FedStaleWeight reframes asynchronous federated learning aggregation as a mechanism design problem, devising a weighting strategy that incentivizes truthful compute speed reporting without favoring faster update-producing agents by upweighting agent updates based on staleness. Leveraging only observed agent update staleness, FedStaleWeight results in more equitable aggregation on a per-agent basis. We both provide theoretical convergence guarantees in the smooth, non-convex setting and empirically compare FedStaleWeight against the commonly used asynchronous FedBuff with gradient averaging, demonstrating how it achieves stronger fairness, expediting convergence to a higher global model accuracy. Finally, we provide an open-source test bench to facilitate exploration of buffered AFL aggregation strategies, fostering further research in asynchronous federated learning paradigms.
Authors: Thushari Hapuarachchi, Jing Lin, Kaiqi Xiong, Mohamed Rahouti, Gitte Ost
Abstract: Automated scraping stands out as a common method for collecting data in deep learning models without the authorization of data owners. Recent studies have begun to tackle the privacy concerns associated with this data collection method. Notable approaches include Deepconfuse, error-minimizing, error-maximizing (also known as adversarial poisoning), Neural Tangent Generalization Attack, synthetic, autoregressive, One-Pixel Shortcut, Self-Ensemble Protection, Entangled Features, Robust Error-Minimizing, Hypocritical, and TensorClog. The data generated by those approaches, called "unlearnable" examples, are prevented "learning" by deep learning models. In this research, we investigate and devise an effective nonlinear transformation framework and conduct extensive experiments to demonstrate that a deep neural network can effectively learn from the data/examples traditionally considered unlearnable produced by the above twelve approaches. The resulting approach improves the ability to break unlearnable data compared to the linear separable technique recently proposed by researchers. Specifically, our extensive experiments show that the improvement ranges from 0.34% to 249.59% for the unlearnable CIFAR10 datasets generated by those twelve data protection approaches, except for One-Pixel Shortcut. Moreover, the proposed framework achieves over 100% improvement of test accuracy for Autoregressive and REM approaches compared to the linear separable technique. Our findings suggest that these approaches are inadequate in preventing unauthorized uses of data in machine learning models. There is an urgent need to develop more robust protection mechanisms that effectively thwart an attacker from accessing data without proper authorization from the owners.
Authors: Rafael Rafailov, Yaswanth Chittepu, Ryan Park, Harshit Sikchi, Joey Hejna, Bradley Knox, Chelsea Finn, Scott Niekum
Abstract: Reinforcement Learning from Human Feedback (RLHF) has been crucial to the recent success of Large Language Models (LLMs), however, it is often a complex and brittle process. In the classical RLHF framework, a reward model is first trained to represent human preferences, which is in turn used by an online reinforcement learning (RL) algorithm to optimize the LLM. A prominent issue with such methods is \emph{reward over-optimization} or \emph{reward hacking}, where performance as measured by the learned proxy reward model increases, but true quality plateaus or even deteriorates. Direct Alignment Algorithms (DDAs) like Direct Preference Optimization have emerged as alternatives to the classical RLHF pipeline by circumventing the reward modeling phase. However, although DAAs do not use a separate proxy reward model, they still commonly deteriorate from over-optimization. While the so-called reward hacking phenomenon is not well-defined for DAAs, we still uncover similar trends: at higher KL budgets, DAA algorithms exhibit similar degradation patterns to their classic RLHF counterparts. In particular, we find that DAA methods deteriorate not only across a wide range of KL budgets but also often before even a single epoch of the dataset is completed. Through extensive empirical experimentation, this work formulates and formalizes the reward over-optimization or hacking problem for DAAs and explores its consequences across objectives, training regimes, and model scales.
Authors: Wentao Guo, Jikai Long, Yimeng Zeng, Zirui Liu, Xinyu Yang, Yide Ran, Jacob R. Gardner, Osbert Bastani, Christopher De Sa, Xiaodong Yu, Beidi Chen, Zhaozhuo Xu
Abstract: Zeroth-order optimization (ZO) is a memory-efficient strategy for fine-tuning Large Language Models using only forward passes. However, the application of ZO fine-tuning in memory-constrained settings such as mobile phones and laptops is still challenging since full precision forward passes are infeasible. In this study, we address this limitation by integrating sparsity and quantization into ZO fine-tuning of LLMs. Specifically, we investigate the feasibility of fine-tuning an extremely small subset of LLM parameters using ZO. This approach allows the majority of un-tuned parameters to be quantized to accommodate the constraint of limited device memory. Our findings reveal that the pre-training process can identify a set of "sensitive parameters" that can guide the ZO fine-tuning of LLMs on downstream tasks. Our results demonstrate that fine-tuning 0.1% sensitive parameters in the LLM with ZO can outperform the full ZO fine-tuning performance, while offering wall-clock time speedup. Additionally, we show that ZO fine-tuning targeting these 0.1% sensitive parameters, combined with 4 bit quantization, enables efficient ZO fine-tuning of an Llama2-7B model on a GPU device with less than 8 GiB of memory and notably reduced latency.
Authors: Khemraj Shukla, Juan Diego Toscano, Zhicheng Wang, Zongren Zou, George Em Karniadakis
Abstract: Kolmogorov-Arnold Networks (KANs) were recently introduced as an alternative representation model to MLP. Herein, we employ KANs to construct physics-informed machine learning models (PIKANs) and deep operator models (DeepOKANs) for solving differential equations for forward and inverse problems. In particular, we compare them with physics-informed neural networks (PINNs) and deep operator networks (DeepONets), which are based on the standard MLP representation. We find that although the original KANs based on the B-splines parameterization lack accuracy and efficiency, modified versions based on low-order orthogonal polynomials have comparable performance to PINNs and DeepONet although they still lack robustness as they may diverge for different random seeds or higher order orthogonal polynomials. We visualize their corresponding loss landscapes and analyze their learning dynamics using information bottleneck theory. Our study follows the FAIR principles so that other researchers can use our benchmarks to further advance this emerging topic.
Authors: Peijie Dong, Lujun Li, Zhenheng Tang, Xiang Liu, Xinglin Pan, Qiang Wang, Xiaowen Chu
Abstract: Despite the remarkable capabilities, Large Language Models (LLMs) face deployment challenges due to their extensive size. Pruning methods drop a subset of weights to accelerate, but many of them require retraining, which is prohibitively expensive and computationally demanding. Recently, post-training pruning approaches introduced novel metrics, enabling the pruning of LLMs without retraining. However, these metrics require the involvement of human experts and tedious trial and error. To efficiently identify superior pruning metrics, we develop an automatic framework for searching symbolic pruning metrics using genetic programming. In particular, we devise an elaborate search space encompassing the existing pruning metrics to discover the potential symbolic pruning metric. We propose an opposing operation simplification strategy to increase the diversity of the population. In this way, Pruner-Zero allows auto-generation of symbolic pruning metrics. Based on the searched results, we explore the correlation between pruning metrics and performance after pruning and summarize some principles. Extensive experiments on LLaMA and LLaMA-2 on language modeling and zero-shot tasks demonstrate that our Pruner-Zero obtains superior performance than SOTA post-training pruning methods. Code at: \url{https://github.com/pprp/Pruner-Zero}.
Authors: Zhenyu Hou, Haozhan Li, Yukuo Cen, Jie Tang, Yuxiao Dong
Abstract: Graph self-supervised learning (SSL) holds considerable promise for mining and learning with graph-structured data. Yet, a significant challenge in graph SSL lies in the feature discrepancy among graphs across different domains. In this work, we aim to pretrain one graph neural network (GNN) on a varied collection of graphs endowed with rich node features and subsequently apply the pretrained GNN to unseen graphs. We present a general GraphAlign method that can be seamlessly integrated into the existing graph SSL framework. To align feature distributions across disparate graphs, GraphAlign designs alignment strategies of feature encoding, normalization, alongside a mixture-of-feature-expert module. Extensive experiments show that GraphAlign empowers existing graph SSL frameworks to pretrain a unified and powerful GNN across multiple graphs, showcasing performance superiority on both in-domain and out-of-domain graphs.
Authors: Charlie Hou, Akshat Shrivastava, Hongyuan Zhan, Rylan Conway, Trang Le, Adithya Sagar, Giulia Fanti, Daniel Lazar
Abstract: On-device training is currently the most common approach for training machine learning (ML) models on private, distributed user data. Despite this, on-device training has several drawbacks: (1) most user devices are too small to train large models on-device, (2) on-device training is communication- and computation-intensive, and (3) on-device training can be difficult to debug and deploy. To address these problems, we propose Private Evolution-Text (PrE-Text), a method for generating differentially private (DP) synthetic textual data. First, we show that across multiple datasets, training small models (models that fit on user devices) with PrE-Text synthetic data outperforms small models trained on-device under practical privacy regimes ($\epsilon=1.29$, $\epsilon=7.58$). We achieve these results while using 9$\times$ fewer rounds, 6$\times$ less client computation per round, and 100$\times$ less communication per round. Second, finetuning large models on PrE-Text's DP synthetic data improves large language model (LLM) performance on private data across the same range of privacy budgets. Altogether, these results suggest that training on DP synthetic data can be a better option than training a model on-device on private distributed data. Code is available at https://github.com/houcharlie/PrE-Text.
Authors: Raeid Saqur, Anastasis Kratsios, Florian Krach, Yannick Limmer, Jacob-Junqi Tian, John Willes, Blanka Horvath, Frank Rudzicz
Abstract: We propose MoE-F -- a formalised mechanism for combining $N$ pre-trained expert Large Language Models (LLMs) in online time-series prediction tasks by adaptively forecasting the best weighting of LLM predictions at every time step. Our mechanism leverages the conditional information in each expert's running performance to forecast the best combination of LLMs for predicting the time series in its next step. Diverging from static (learned) Mixture of Experts (MoE) methods, MoE-F employs time-adaptive stochastic filtering techniques to combine experts. By framing the expert selection problem as a finite state-space, continuous-time Hidden Markov model (HMM), we can leverage the Wohman-Shiryaev filter. Our approach first constructs $N$ parallel filters corresponding to each of the $N$ individual LLMs. Each filter proposes its best combination of LLMs, given the information that they have access to. Subsequently, the $N$ filter outputs are aggregated to optimize a lower bound for the loss of the aggregated LLMs, which can be optimized in closed-form, thus generating our ensemble predictor. Our contributions here are: (I) the MoE-F algorithm -- deployable as a plug-and-play filtering harness, (II) theoretical optimality guarantees of the proposed filtering-based gating algorithm, and (III) empirical evaluation and ablative results using state of the art foundational and MoE LLMs on a real-world Financial Market Movement task where MoE-F attains a remarkable 17% absolute and 48.5% relative F1 measure improvement over the next best performing individual LLM expert.
Authors: Yucheng Wu, Liyue Chen, Yu Cheng, Shuai Chen, Jinyu Xu, Leye Wang
Abstract: Learning representations of user behavior sequences is crucial for various online services, such as online fraudulent transaction detection mechanisms. Graph Neural Networks (GNNs) have been extensively applied to model sequence relationships, and extract information from similar sequences. While user behavior sequence data volume is usually huge for online applications, directly applying GNN models may lead to substantial computational overhead during both the training and inference stages and make it challenging to meet real-time requirements for online services. In this paper, we leverage graph compression techniques to alleviate the efficiency issue. Specifically, we propose a novel unified framework called ECSeq, to introduce graph compression techniques into relation modeling for user sequence representation learning. The key module of ECSeq is sequence relation modeling, which explores relationships among sequences to enhance sequence representation learning, and employs graph compression algorithms to achieve high efficiency and scalability. ECSeq also exhibits plug-and-play characteristics, seamlessly augmenting pre-trained sequence representation models without modifications. Empirical experiments on both sequence classification and regression tasks demonstrate the effectiveness of ECSeq. Specifically, with an additional training time of tens of seconds in total on 100,000+ sequences and inference time preserved within $10^{-4}$ seconds/sample, ECSeq improves the prediction R@P$_{0.9}$ of the widely used LSTM by $\sim 5\%$.
Authors: Yang Chen, Ce Zhu, Jiani Liu, Yipeng Liu
Abstract: Additive models can be used for interpretable machine learning for their clarity and simplicity. However, In the classical models for high-order data, the vectorization operation disrupts the data structure, which may lead to degenerated accuracy and increased computational complexity. To deal with these problems, we propose the tensor polynomial addition model (TPAM). It retains the multidimensional structure information of high-order inputs with tensor representation. The model parameter compression is achieved using a hierarchical and low-order symmetric tensor approximation. In this way, complex high-order feature interactions can be captured with fewer parameters. Moreover, The TPAM preserves the inherent interpretability of additive models, facilitating transparent decision-making and the extraction of meaningful feature values. Additionally, leveraging TPAM's transparency and ability to handle higher-order features, it is used as a post-processing module for other interpretation models by introducing two variants for class activation maps. Experimental results on a series of datasets demonstrate that TPAM can enhance accuracy by up to 30\%, and compression rate by up to 5 times, while maintaining a good interpretability.
Authors: Shahaf Bassan, Guy Amir, Guy Katz
Abstract: The local and global interpretability of various ML models has been studied extensively in recent years. However, despite significant progress in the field, many known results remain informal or lack sufficient mathematical rigor. We propose a framework for bridging this gap, by using computational complexity theory to assess local and global perspectives of interpreting ML models. We begin by proposing proofs for two novel insights that are essential for our analysis: (1) a duality between local and global forms of explanations; and (2) the inherent uniqueness of certain global explanation forms. We then use these insights to evaluate the complexity of computing explanations, across three model types representing the extremes of the interpretability spectrum: (1) linear models; (2) decision trees; and (3) neural networks. Our findings offer insights into both the local and global interpretability of these models. For instance, under standard complexity assumptions such as P != NP, we prove that selecting global sufficient subsets in linear models is computationally harder than selecting local subsets. Interestingly, with neural networks and decision trees, the opposite is true: it is harder to carry out this task locally than globally. We believe that our findings demonstrate how examining explainability through a computational complexity lens can help us develop a more rigorous grasp of the inherent interpretability of ML models.
Authors: Wooseong Jeong, Kuk-Jin Yoon
Abstract: The goal of multi-task learning is to learn diverse tasks within a single unified network. As each task has its own unique objective function, conflicts emerge during training, resulting in negative transfer among them. Earlier research identified these conflicting gradients in shared parameters between tasks and attempted to realign them in the same direction. However, we prove that such optimization strategies lead to sub-optimal Pareto solutions due to their inability to accurately determine the individual contributions of each parameter across various tasks. In this paper, we propose the concept of task priority to evaluate parameter contributions across different tasks. To learn task priority, we identify the type of connections related to links between parameters influenced by task-specific losses during backpropagation. The strength of connections is gauged by the magnitude of parameters to determine task priority. Based on these, we present a new method named connection strength-based optimization for multi-task learning which consists of two phases. The first phase learns the task priority within the network, while the second phase modifies the gradients while upholding this priority. This ultimately leads to finding new Pareto optimal solutions for multiple tasks. Through extensive experiments, we show that our approach greatly enhances multi-task performance in comparison to earlier gradient manipulation methods.
Authors: Michael Scholkemper, Xinyi Wu, Ali Jadbabaie, Michael Schaub
Abstract: Residual connections and normalization layers have become standard design choices for graph neural networks (GNNs), and were proposed as solutions to the mitigate the oversmoothing problem in GNNs. However, how exactly these methods help alleviate the oversmoothing problem from a theoretical perspective is not well understood. In this work, we provide a formal and precise characterization of (linearized) GNNs with residual connections and normalization layers. We establish that (a) for residual connections, the incorporation of the initial features at each layer can prevent the signal from becoming too smooth, and determines the subspace of possible node representations; (b) batch normalization prevents a complete collapse of the output embedding space to a one-dimensional subspace through the individual rescaling of each column of the feature matrix. This results in the convergence of node representations to the top-$k$ eigenspace of the message-passing operator; (c) moreover, we show that the centering step of a normalization layer -- which can be understood as a projection -- alters the graph signal in message-passing in such a way that relevant information can become harder to extract. We therefore introduce a novel, principled normalization layer called GraphNormv2 in which the centering step is learned such that it does not distort the original graph signal in an undesirable way. Experimental results confirm the effectiveness of our method.
Authors: Andr\'e Artelt, Barbara Hammer
Abstract: EXplainable AI (XAI) constitutes a popular method to analyze the reasoning of AI systems by explaining their decision-making, e.g. providing a counterfactual explanation of how to achieve recourse. However, in cases such as unexpected explanations, the user might be interested in learning about the cause of this explanation -- e.g. properties of the utilized training data that are responsible for the observed explanation. Under the umbrella of data valuation, first approaches have been proposed that estimate the influence of data samples on a given model. In this work, we take a slightly different stance, as we are interested in the influence of single samples on a model explanation rather than the model itself. Hence, we propose the novel problem of identifying training data samples that have a high influence on a given explanation (or related quantity) and investigate the particular case of differences in the cost of the recourse between protected groups. For this, we propose an algorithm that identifies such influential training samples.
Authors: Riccardo Poiani, R\'emy Degenne, Emilie Kaufmann, Alberto Maria Metelli, Marcello Restelli
Abstract: In bandit best-arm identification, an algorithm is tasked with finding the arm with highest mean reward with a specified accuracy as fast as possible. We study multi-fidelity best-arm identification, in which the algorithm can choose to sample an arm at a lower fidelity (less accurate mean estimate) for a lower cost. Several methods have been proposed for tackling this problem, but their optimality remain elusive, notably due to loose lower bounds on the total cost needed to identify the best arm. Our first contribution is a tight, instance-dependent lower bound on the cost complexity. The study of the optimization problem featured in the lower bound provides new insights to devise computationally efficient algorithms, and leads us to propose a gradient-based approach with asymptotically optimal cost complexity. We demonstrate the benefits of the new algorithm compared to existing methods in experiments. Our theoretical and empirical findings also shed light on an intriguing concept of optimal fidelity for each arm.
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 intracranial neural recordings at scale, unlocking the benefits of representation learning for a key neuroscience recording modality. The Population Transformer (PopT) lowers the amount of data required for decoding experiments, while increasing accuracy, even on never-before-seen subjects and tasks. We address two key challenges in developing PopT: sparse electrode distribution and varying electrode location across patients. PopT stacks on top of pretrained representations and enhances downstream tasks by enabling learned aggregation of multiple spatially-sparse data channels. Beyond decoding, we interpret the pretrained PopT and fine-tuned models to show how it can be used to provide neuroscience insights learned from massive amounts of data. We release a pretrained PopT to enable off-the-shelf improvements in multi-channel intracranial data decoding and interpretability, and code is available at https://github.com/czlwang/PopulationTransformer.
Authors: Zihan Luo, Hong Huang, Yongkang Zhou, Jiping Zhang, Nuo Chen
Abstract: Despite the remarkable capabilities demonstrated by Graph Neural Networks (GNNs) in graph-related tasks, recent research has revealed the fairness vulnerabilities in GNNs when facing malicious adversarial attacks. However, all existing fairness attacks require manipulating the connectivity between existing nodes, which may be prohibited in reality. To this end, we introduce a Node Injection-based Fairness Attack (NIFA), exploring the vulnerabilities of GNN fairness in such a more realistic setting. In detail, NIFA first designs two insightful principles for node injection operations, namely the uncertainty-maximization principle and homophily-increase principle, and then optimizes injected nodes' feature matrix to further ensure the effectiveness of fairness attacks. Comprehensive experiments on three real-world datasets consistently demonstrate that NIFA can significantly undermine the fairness of mainstream GNNs, even including fairness-aware GNNs, by injecting merely 1% of nodes. We sincerely hope that our work can stimulate increasing attention from researchers on the vulnerability of GNN fairness, and encourage the development of corresponding defense mechanisms.
Authors: Hoyong Choi, Nohyun Ki, Hye Won Chung
Abstract: Data subset selection aims to find a smaller yet informative subset of a large dataset that can approximate the full-dataset training, addressing challenges associated with training neural networks on large-scale datasets. However, existing methods tend to specialize in either high or low selection ratio regimes, lacking a universal approach that consistently achieves competitive performance across a broad range of selection ratios. We introduce a universal and efficient data subset selection method, Best Window Selection (BWS), by proposing a method to choose the best window subset from samples ordered based on their difficulty scores. This approach offers flexibility by allowing the choice of window intervals that span from easy to difficult samples. Furthermore, we provide an efficient mechanism for selecting the best window subset by evaluating its quality using kernel ridge regression. Our experimental results demonstrate the superior performance of BWS compared to other baselines across a broad range of selection ratios over datasets, including CIFAR-10/100 and ImageNet, and the scenarios involving training from random initialization or fine-tuning of pre-trained models.
Authors: Martino Ciaperoni, Han Xiao, Aristides Gionis
Abstract: Today, as increasingly complex predictive models are developed, simple rule sets remain a crucial tool to obtain interpretable predictions and drive high-stakes decision making. However, a single rule set provides a partial representation of a learning task. An emerging paradigm in interpretable machine learning aims at exploring the Rashomon set of all models exhibiting near-optimal performance. Existing work on Rashomon-set exploration focuses on exhaustive search of the Rashomon set for particular classes of models, which can be a computationally challenging task. On the other hand, exhaustive enumeration leads to redundancy that often is not necessary, and a representative sample or an estimate of the size of the Rashomon set is sufficient for many applications. In this work, we propose, for the first time, efficient methods to explore the Rashomon set of rule set models with or without exhaustive search. Extensive experiments demonstrate the effectiveness of the proposed methods in a variety of scenarios.
Authors: Shihori Koyama, Daisuke Inoue, Hiroaki Yoshida, Kazuyuki Aihara, Gouhei Tanaka
Abstract: Collecting time series data spatially distributed in many locations is often important for analyzing climate change and its impacts on ecosystems. However, comprehensive spatial data collection is not always feasible, requiring us to predict climate variables at some locations. This study focuses on a prediction of climatic elements, specifically near-surface temperature and pressure, at a target location apart from a data observation point. Our approach uses two prediction methods: reservoir computing (RC), known as a machine learning framework with low computational requirements, and vector autoregression models (VAR), recognized as a statistical method for analyzing time series data. Our results show that the accuracy of the predictions degrades with the distance between the observation and target locations. We quantitatively estimate the distance in which effective predictions are possible. We also find that in the context of climate data, a geographical distance is associated with data correlation, and a strong data correlation significantly improves the prediction accuracy with RC. In particular, RC outperforms VAR in predicting highly correlated data within the predictive range. These findings suggest that machine learning-based methods can be used more effectively to predict climatic elements in remote locations by assessing the distance to them from the data observation point in advance. Our study on low-cost and accurate prediction of climate variables has significant value for climate change strategies.
Authors: Dacao Zhang, Kun Zhang, Le Wu, Mi Tian, Richang Hong, Meng Wang
Abstract: Cognitive Diagnosis~(CD), which leverages students and exercise data to predict students' proficiency levels on different knowledge concepts, is one of fundamental components in Intelligent Education. Due to the scarcity of student-exercise interaction data, most existing methods focus on making the best use of available data, such as exercise content and student information~(e.g., educational context). Despite the great progress, the abuse of student sensitive information has not been paid enough attention. Due to the important position of CD in Intelligent Education, employing sensitive information when making diagnosis predictions will cause serious social issues. Moreover, data-driven neural networks are easily misled by the shortcut between input data and output prediction, exacerbating this problem. Therefore, it is crucial to eliminate the negative impact of sensitive information in CD models. In response, we argue that sensitive attributes of students can also provide useful information, and only the shortcuts directly related to the sensitive information should be eliminated from the diagnosis process. Thus, we employ causal reasoning and design a novel Path-Specific Causal Reasoning Framework (PSCRF) to achieve this goal. Specifically, we first leverage an encoder to extract features and generate embeddings for general information and sensitive information of students. Then, we design a novel attribute-oriented predictor to decouple the sensitive attributes, in which fairness-related sensitive features will be eliminated and other useful information will be retained. Finally, we designed a multi-factor constraint to ensure the performance of fairness and diagnosis performance simultaneously. Extensive experiments over real-world datasets (e.g., PISA dataset) demonstrate the effectiveness of our proposed PSCRF.
Authors: Qiang Nie, Weifu Fu, Yuhuan Lin, Jialin Li, Yifeng Zhou, Yong Liu, Lei Zhu, Chengjie Wang
Abstract: Instance-incremental learning (IIL) focuses on learning continually with data of the same classes. Compared to class-incremental learning (CIL), the IIL is seldom explored because IIL suffers less from catastrophic forgetting (CF). However, besides retaining knowledge, in real-world deployment scenarios where the class space is always predefined, continual and cost-effective model promotion with the potential unavailability of previous data is a more essential demand. Therefore, we first define a new and more practical IIL setting as promoting the model's performance besides resisting CF with only new observations. Two issues have to be tackled in the new IIL setting: 1) the notorious catastrophic forgetting because of no access to old data, and 2) broadening the existing decision boundary to new observations because of concept drift. To tackle these problems, our key insight is to moderately broaden the decision boundary to fail cases while retain old boundary. Hence, we propose a novel decision boundary-aware distillation method with consolidating knowledge to teacher to ease the student learning new knowledge. We also establish the benchmarks on existing datasets Cifar-100 and ImageNet. Notably, extensive experiments demonstrate that the teacher model can be a better incremental learner than the student model, which overturns previous knowledge distillation-based methods treating student as the main role.
Authors: Lei Chen, Joan Bruna, Alberto Bietti
Abstract: In addition to the ability to generate fluent text in various languages, large language models have been successful at tasks that involve basic forms of logical "reasoning" over their context. Recent work found that selectively removing certain components from weight matrices in pre-trained models can improve such reasoning capabilities. We investigate this phenomenon further by carefully studying how certain global associations tend to be stored in specific weight components or Transformer blocks, in particular feed-forward layers. Such associations may hurt predictions in reasoning tasks, and removing the corresponding components may then improve performance. We analyze how this arises during training, both empirically and theoretically, on a two-layer Transformer trained on a basic reasoning task with noise, a toy associative memory model, and on the Pythia family of pre-trained models tested on simple reasoning tasks.
Authors: Ashok Vardhan Makkuva, Marco Bondaschi, Chanakya Ekbote, Adway Girish, Alliot Nagle, Hyeji Kim, Michael Gastpar
Abstract: In recent years, transformer-based models have revolutionized deep learning, particularly in sequence modeling. To better understand this phenomenon, there is a growing interest in using Markov input processes to study transformers. However, our current understanding in this regard remains limited with many fundamental questions about how transformers learn Markov chains still unanswered. In this paper, we address this by focusing on first-order Markov chains and single-layer transformers, providing a comprehensive characterization of the learning dynamics in this context. Specifically, we prove that transformer parameters trained on next-token prediction loss can either converge to global or local minima, contingent on the initialization and the Markovian data properties, and we characterize the precise conditions under which this occurs. To the best of our knowledge, this is the first result of its kind highlighting the role of initialization. We further demonstrate that our theoretical findings are corroborated by empirical evidence. Based on these insights, we provide guidelines for the initialization of transformer parameters and demonstrate their effectiveness. Finally, we outline several open problems in this arena. Code is available at: \url{https://anonymous.4open.science/r/Local-to-Global-C70B/}.
URLs: https://anonymous.4open.science/r/Local-to-Global-C70B/
Authors: Kahou Tam, Kewei Xu, Li Li, Huazhu Fu
Abstract: Federated Learning (FL) has evolved as a powerful tool for collaborative model training across multiple entities, ensuring data privacy in sensitive sectors such as healthcare and finance. However, the introduction of the Right to Be Forgotten (RTBF) poses new challenges, necessitating federated unlearning to delete data without full model retraining. Traditional FL unlearning methods, not originally designed with domain specificity in mind, inadequately address the complexities of multi-domain scenarios, often affecting the accuracy of models in non-targeted domains or leading to uniform forgetting across all domains. Our work presents the first comprehensive empirical study on Federated Domain Unlearning, analyzing the characteristics and challenges of current techniques in multi-domain contexts. We uncover that these methods falter, particularly because they neglect the nuanced influences of domain-specific data, which can lead to significant performance degradation and inaccurate model behavior. Our findings reveal that unlearning disproportionately affects the model's deeper layers, erasing critical representational subspaces acquired during earlier training phases. In response, we propose novel evaluation methodologies tailored for Federated Domain Unlearning, aiming to accurately assess and verify domain-specific data erasure without compromising the model's overall integrity and performance. This investigation not only highlights the urgent need for domain-centric unlearning strategies in FL but also sets a new precedent for evaluating and implementing these techniques effectively.
Authors: Alan A. Lahoud, Erik Schaffernicht, Johannes A. Stork
Abstract: Mathematical solvers use parametrized Optimization Problems (OPs) as inputs to yield optimal decisions. In many real-world settings, some of these parameters are unknown or uncertain. Recent research focuses on predicting the value of these unknown parameters using available contextual features, aiming to decrease decision regret by adopting end-to-end learning approaches. However, these approaches disregard prediction uncertainty and therefore make the mathematical solver susceptible to provide erroneous decisions in case of low-confidence predictions. We propose a novel framework that models prediction uncertainty with Bayesian Neural Networks (BNNs) and propagates this uncertainty into the mathematical solver with a Stochastic Programming technique. The differentiable nature of BNNs and differentiable mathematical solvers allow for two different learning approaches: In the Decoupled learning approach, we update the BNN weights to increase the quality of the predictions' distribution of the OP parameters, while in the Combined learning approach, we update the weights aiming to directly minimize the expected OP's cost function in a stochastic end-to-end fashion. We do an extensive evaluation using synthetic data with various noise properties and a real dataset, showing that decisions regret are generally lower (better) with both proposed methods.
Authors: Tingjia Shen, Hao Wang, Jiaqing Zhang, Sirui Zhao, Liangyue Li, Zulong Chen, Defu Lian, Enhong Chen
Abstract: Cross-Domain Sequential Recommendation (CDSR) aims to mine and transfer users' sequential preferences across different domains to alleviate the long-standing cold-start issue. Traditional CDSR models capture collaborative information through user and item modeling while overlooking valuable semantic information. Recently, Large Language Model (LLM) has demonstrated powerful semantic reasoning capabilities, motivating us to introduce them to better capture semantic information. However, introducing LLMs to CDSR is non-trivial due to two crucial issues: seamless information integration and domain-specific generation. To this end, we propose a novel framework named URLLM, which aims to improve the CDSR performance by exploring the User Retrieval approach and domain grounding on LLM simultaneously. Specifically, we first present a novel dual-graph sequential model to capture the diverse information, along with an alignment and contrastive learning method to facilitate domain knowledge transfer. Subsequently, a user retrieve-generation model is adopted to seamlessly integrate the structural information into LLM, fully harnessing its emergent inferencing ability. Furthermore, we propose a domain-specific strategy and a refinement module to prevent out-of-domain generation. Extensive experiments on Amazon demonstrated the information integration and domain-specific generation ability of URLLM in comparison to state-of-the-art baselines. Our code is available at https://github.com/TingJShen/URLLM
Authors: Shuqi He, Jun Zhuang, Ding Wang, Luyao Peng, Jun Song
Abstract: Graph neural networks (GNNs) have been extensively employed in node classification. Nevertheless, recent studies indicate that GNNs are vulnerable to topological perturbations, such as adversarial attacks and edge disruptions. Considerable efforts have been devoted to mitigating these challenges. For example, pioneering Bayesian methodologies, including GraphSS and LlnDT, incorporate Bayesian label transitions and topology-based label sampling to strengthen the robustness of GNNs. However, GraphSS is hindered by slow convergence, while LlnDT faces challenges in sparse graphs. To overcome these limitations, we propose a novel label inference framework, TraTopo, which combines topology-driven label propagation, Bayesian label transitions, and link analysis via random walks. TraTopo significantly surpasses its predecessors on sparse graphs by utilizing random walk sampling, specifically targeting isolated nodes for link prediction, thus enhancing its effectiveness in topological sampling contexts. Additionally, TraTopo employs a shortest-path strategy to refine link prediction, thereby reducing predictive overhead and improving label inference accuracy. Empirical evaluations highlight TraTopo's superiority in node classification, significantly exceeding contemporary GCN models in accuracy.
Authors: Lorenzo Sciandra, Roberto Esposito, Andrea Cesare Grosso, Laura Sacerdote, Cristina Zucca
Abstract: This article demonstrates the effectiveness of employing a deep learning model in an optimization pipeline. Specifically, in a generic exact algorithm for a NP problem, multiple heuristic criteria are usually used to guide the search of the optimum within the set of all feasible solutions. In this context, neural networks can be leveraged to rapidly acquire valuable information, enabling the identification of a more expedient path in this vast space. So, after the explanation of the tackled traveling salesman problem, the implemented branch and bound for its classical resolution is described. This algorithm is then compared with its hybrid version termed "graph convolutional branch and bound" that integrates the previous branch and bound with a graph convolutional neural network. The empirical results obtained highlight the efficacy of this approach, leading to conclusive findings and suggesting potential directions for future research.
Authors: Bo Xia, Yilun Kong, Yongzhe Chang, Bo Yuan, Zhiheng Li, Xueqian Wang, Bin Liang
Abstract: Classic reinforcement learning (RL) frequently confronts challenges in tasks involving delays, which cause a mismatch between received observations and subsequent actions, thereby deviating from the Markov assumption. Existing methods usually tackle this issue with end-to-end solutions using state augmentation. However, these black-box approaches often involve incomprehensible processes and redundant information in the information states, causing instability and potentially undermining the overall performance. To alleviate the delay challenges in RL, we propose $\textbf{DEER (Delay-resilient Encoder-Enhanced RL)}$, a framework designed to effectively enhance the interpretability and address the random delay issues. DEER employs a pretrained encoder to map delayed states, along with their variable-length past action sequences resulting from different delays, into hidden states, which is trained on delay-free environment datasets. In a variety of delayed scenarios, the trained encoder can seamlessly integrate with standard RL algorithms without requiring additional modifications and enhance the delay-solving capability by simply adapting the input dimension of the original algorithms. We evaluate DEER through extensive experiments on Gym and Mujoco environments. The results confirm that DEER is superior to state-of-the-art RL algorithms in both constant and random delay settings.
Authors: Hatem Helal, Andrew Fitzgibbon
Abstract: Electronic structure simulation (ESS) has been used for decades to provide quantitative scientific insights on an atomistic scale, enabling advances in chemistry, biology, and materials science, among other disciplines. Following standard practice in scientific computing, the software packages driving these studies have been implemented in compiled languages such as FORTRAN and C. However, the recent introduction of machine learning (ML) into these domains has meant that ML models must be coded in these languages, or that complex software bridges have to be built between ML models in Python and these large compiled software systems. This is in contrast with recent progress in modern ML frameworks which aim to optimise both ease of use and high performance by harnessing hardware acceleration of tensor programs defined in Python. We introduce MESS: a modern electronic structure simulation package implemented in JAX; porting the ESS code to the ML world. We outline the costs and benefits of following the software development practices used in ML for this important scientific workload. MESS shows significant speedups n widely available hardware accelerators and simultaneously opens a clear pathway towards combining ESS with ML. MESS is available at https://github.com/graphcore-research/mess.
Authors: Jerry Yao-Chieh Hu, Maojiang Su, En-Jui Kuo, Zhao Song, Han Liu
Abstract: We study the computational limits of Low-Rank Adaptation (LoRA) update for finetuning transformer-based models using fine-grained complexity theory. Our key observation is that the existence of low-rank decompositions within the gradient computation of LoRA adaptation leads to possible algorithmic speedup. This allows us to (i) identify a phase transition behavior and (ii) prove the existence of nearly linear algorithms by controlling the LoRA update computation term by term, assuming the Strong Exponential Time Hypothesis (SETH). For the former, we identify a sharp transition in the efficiency of all possible rank-$r$ LoRA update algorithms for transformers, based on specific norms resulting from the multiplications of the input sequence $\mathbf{X}$, pretrained weights $\mathbf{W^\star}$, and adapter matrices $\alpha \mathbf{B} \mathbf{A} / r$. Specifically, we derive a shared upper bound threshold for such norms and show that efficient (sub-quadratic) approximation algorithms of LoRA exist only below this threshold. For the latter, we prove the existence of nearly linear approximation algorithms for LoRA adaptation by utilizing the hierarchical low-rank structures of LoRA gradients and approximating the gradients with a series of chained low-rank approximations. To showcase our theory, we consider two practical scenarios: partial (e.g., only $\mathbf{W}_V$ and $\mathbf{W}_Q$) and full adaptations (e.g., $\mathbf{W}_Q$, $\mathbf{W}_V$, and $\mathbf{W}_K$) of weights in attention heads.
Authors: Sanghyun Lee, Chanyoung Park
Abstract: The real-world traffic networks undergo expansion through the installation of new sensors, implying that the traffic patterns continually evolve over time. Incrementally training a model on the newly added sensors would make the model forget the past knowledge, i.e., catastrophic forgetting, while retraining the model on the entire network to capture these changes is highly inefficient. To address these challenges, we propose a novel Traffic Forecasting Mixture of Experts (TFMoE) for traffic forecasting under evolving networks. The main idea is to segment the traffic flow into multiple homogeneous groups, and assign an expert model responsible for a specific group. This allows each expert model to concentrate on learning and adapting to a specific set of patterns, while minimizing interference between the experts during training, thereby preventing the dilution or replacement of prior knowledge, which is a major cause of catastrophic forgetting. Through extensive experiments on a real-world long-term streaming network dataset, PEMSD3-Stream, we demonstrate the effectiveness and efficiency of TFMoE. Our results showcase superior performance and resilience in the face of catastrophic forgetting, underscoring the effectiveness of our approach in dealing with continual learning for traffic flow forecasting in long-term streaming networks.
Authors: Sushant Agarwal, Amit Deshpande
Abstract: Fair classification and fair representation learning are two important problems in supervised and unsupervised fair machine learning, respectively. Fair classification asks for a classifier that maximizes accuracy on a given data distribution subject to fairness constraints. Fair representation maps a given data distribution over the original feature space to a distribution over a new representation space such that all classifiers over the representation satisfy fairness. In this paper, we examine the power of randomization in both these problems to minimize the loss of accuracy that results when we impose fairness constraints. Previous work on fair classification has characterized the optimal fair classifiers on a given data distribution that maximize accuracy subject to fairness constraints, e.g., Demographic Parity (DP), Equal Opportunity (EO), and Predictive Equality (PE). We refine these characterizations to demonstrate when the optimal randomized fair classifiers can surpass their deterministic counterparts in accuracy. We also show how the optimal randomized fair classifier that we characterize can be obtained as a solution to a convex optimization problem. Recent work has provided techniques to construct fair representations for a given data distribution such that any classifier over this representation satisfies DP. However, the classifiers on these fair representations either come with no or weak accuracy guarantees when compared to the optimal fair classifier on the original data distribution. Extending our ideas for randomized fair classification, we improve on these works, and construct DP-fair, EO-fair, and PE-fair representations that have provably optimal accuracy and suffer no accuracy loss compared to the optimal DP-fair, EO-fair, and PE-fair classifiers respectively on the original data distribution.
Authors: Veljko Kovac (University of Amsterdam), Erik J. Bekkers (University of Amsterdam), Pietro Li\`o (Department of Computer Science and Technology, University of Cambridge), Floor Eijkelboom (University of Amsterdam)
Abstract: This paper introduces E(n) Equivariant Message Passing Cellular Networks (EMPCNs), an extension of E(n) Equivariant Graph Neural Networks to CW-complexes. Our approach addresses two aspects of geometric message passing networks: 1) enhancing their expressiveness by incorporating arbitrary cells, and 2) achieving this in a computationally efficient way with a decoupled EMPCNs technique. We demonstrate that EMPCNs achieve close to state-of-the-art performance on multiple tasks without the need for steerability, including many-body predictions and motion capture. Moreover, ablation studies confirm that decoupled EMPCNs exhibit stronger generalization capabilities than their non-topologically informed counterparts. These findings show that EMPCNs can be used as a scalable and expressive framework for higher-order message passing in geometric and topological graphs
Authors: Luis M\"uller, Christopher Morris
Abstract: Graph neural network architectures aligned with the $k$-dimensional Weisfeiler--Leman ($k$-WL) hierarchy offer theoretically well-understood expressive power. However, these architectures often fail to deliver state-of-the-art predictive performance on real-world graphs, limiting their practical utility. While recent works aligning graph transformer architectures with the $k$-WL hierarchy have shown promising empirical results, employing transformers for higher orders of $k$ remains challenging due to a prohibitive runtime and memory complexity of self-attention as well as impractical architectural assumptions, such as an infeasible number of attention heads. Here, we advance the alignment of transformers with the $k$-WL hierarchy, showing stronger expressivity results for each $k$, making them more feasible in practice. In addition, we develop a theoretical framework that allows the study of established positional encodings such as Laplacian PEs and SPE. We evaluate our transformers on the large-scale PCQM4Mv2 dataset, showing competitive predictive performance with the state-of-the-art and demonstrating strong downstream performance when fine-tuning them on small-scale molecular datasets. Our code is available at https://github.com/luis-mueller/wl-transformers.
Authors: Chengyi Cai, Zesheng Ye, Lei Feng, Jianzhong Qi, Feng Liu
Abstract: Visual reprogramming (VR) is a prompting technique that aims to re-purpose a pre-trained model (e.g., a classifier on ImageNet) to target tasks (e.g., medical data prediction) by learning a small-scale pattern added into input images instead of tuning considerable parameters within the model. The location of the pattern within input samples is usually determined by a pre-defined mask shared across all samples. In this paper, we show that the shared mask potentially limits VR's generalization and increases its approximation error due to the lack of sample-level adaptation. Motivated by this finding, we design a new framework for VR called sample-specific multi-channel masks (SMM). Specifically, SMM employs a lightweight ConvNet and patch-wise interpolation to generate sample-specific three-channel masks instead of a shared and pre-defined mask. Since we generate different masks for individual samples, SMM is theoretically shown to reduce approximation error for the target tasks compared with existing state-of-the-art VR methods. We also empirically demonstrate its performance gain on both ResNet and ViT. The success of SMM further highlights the broader applicability of VR in leveraging the latent knowledge of pre-trained models for various target tasks. Our code is available at https://github.com/tmlr-group/SMM.
Authors: Marvin Schmitt, Paul-Christian B\"urkner, Ullrich K\"othe, Stefan T. Radev
Abstract: Recent advances in probabilistic deep learning enable efficient amortized Bayesian inference in settings where the likelihood function is only implicitly defined by a simulation program (simulation-based inference; SBI). But how faithful is such inference if the simulation represents reality somewhat inaccurately, that is, if the true system behavior at test time deviates from the one seen during training? We conceptualize the types of such model misspecification arising in SBI and systematically investigate how the performance of neural posterior approximators gradually deteriorates as a consequence, making inference results less and less trustworthy. To notify users about this problem, we propose a new misspecification measure that can be trained in an unsupervised fashion (i.e., without training data from the true distribution) and reliably detects model misspecification at test time. Our experiments clearly demonstrate the utility of our new measure both on toy examples with an analytical ground-truth and on representative scientific tasks in cell biology, cognitive decision making, disease outbreak dynamics, and computer vision. We show how the proposed misspecification test warns users about suspicious outputs, raises an alarm when predictions are not trustworthy, and guides model designers in their search for better simulators.
Authors: Louis Chislett, Louis JM Aslett, Alisha R Davies, Catalina A Vallejos, James Liley
Abstract: Clinical prediction models are statistical or machine learning models used to quantify the risk of a certain health outcome using patient data. These can then inform potential interventions on patients, causing an effect called performative prediction: predictions inform interventions which influence the outcome they were trying to predict, leading to a potential underestimation of risk in some patients if a model is updated on this data. One suggested resolution to this is the use of hold-out sets, in which a set of patients do not receive model derived risk scores, such that a model can be safely retrained. We present an overview of clinical and research ethics regarding potential implementation of hold-out sets for clinical prediction models in health settings. We focus on the ethical principles of beneficence, non-maleficence, autonomy and justice. We also discuss informed consent, clinical equipoise, and truth-telling. We present illustrative cases of potential hold-out set implementations and discuss statistical issues arising from different hold-out set sampling methods. We also discuss differences between hold-out sets and randomised control trials, in terms of ethics and statistical issues. Finally, we give practical recommendations for researchers interested in the use hold-out sets for clinical prediction models.
Authors: Yogesh Verma, Amauri H Souza, Vikas Garg
Abstract: Topological Neural Networks (TNNs) incorporate higher-order relational information beyond pairwise interactions, enabling richer representations than Graph Neural Networks (GNNs). Concurrently, topological descriptors based on persistent homology (PH) are being increasingly employed to augment the GNNs. We investigate the benefits of integrating these two paradigms. Specifically, we introduce TopNets as a broad framework that subsumes and unifies various methods in the intersection of GNNs/TNNs and PH such as (generalizations of) RePHINE and TOGL. TopNets can also be readily adapted to handle (symmetries in) geometric complexes, extending the scope of TNNs and PH to spatial settings. Theoretically, we show that PH descriptors can provably enhance the expressivity of simplicial message-passing networks. Empirically, (continuous and E(n)-equivariant extensions of) TopNets achieve strong performance across diverse tasks, including antibody design, molecular dynamics simulation, and drug property prediction.
Authors: Chenhao Si, Ming Yan
Abstract: We propose a new physics-informed neural network framework, IDPINN, based on the enhancement of initialization and domain decomposition to improve prediction accuracy. We train a PINN using a small dataset to obtain an initial network structure, including the weighted matrix and bias, which initializes the PINN for each subdomain. Moreover, we leverage the smoothness condition on the interface to enhance the prediction performance. We numerically evaluated it on several forward problems and demonstrated the benefits of IDPINN in terms of accuracy.
Authors: Amir Mohammad Karimi Mamaghan, Panagiotis Tigas, Karl Henrik Johansson, Yarin Gal, Yashas Annadani, Stefan Bauer
Abstract: Representing uncertainty in causal discovery is a crucial component for experimental design, and more broadly, for safe and reliable causal decision making. Bayesian Causal Discovery (BCD) offers a principled approach to encapsulating this uncertainty. Unlike non-Bayesian causal discovery, which relies on a single estimated causal graph and model parameters for assessment, evaluating BCD presents challenges due to the nature of its inferred quantity - the posterior distribution. As a result, the research community has proposed various metrics to assess the quality of the approximate posterior. However, there is, to date, no consensus on the most suitable metric(s) for evaluation. In this work, we reexamine this question by dissecting various metrics and understanding their limitations. Through extensive empirical evaluation, we find that many existing metrics fail to exhibit a strong correlation with the quality of approximation to the true posterior, especially in scenarios with low sample sizes where BCD is most desirable. We highlight the suitability (or lack thereof) of these metrics under two distinct factors: the identifiability of the underlying causal model and the quantity of available data. Both factors affect the entropy of the true posterior, indicating that the current metrics are less fitting in settings of higher entropy. Our findings underline the importance of a more nuanced evaluation of new methods by taking into account the nature of the true posterior, as well as guide and motivate the development of new evaluation procedures for this challenge.
Authors: Josuan Calderon, Gordon J. Berman
Abstract: Most approaches for assessing causality in complex dynamical systems fail when the interactions between variables are inherently non-linear and non-stationary. Here we introduce Temporal Autoencoders for Causal Inference (TACI), a methodology that combines a new surrogate data metric for assessing causal interactions with a novel two-headed machine learning architecture to identify and measure the direction and strength of time-varying causal interactions. Through tests on both synthetic and real-world datasets, we demonstrate TACI's ability to accurately quantify dynamic causal interactions across a variety of systems. Our findings display the method's effectiveness compared to existing approaches and also highlight our approach's potential to build a deeper understanding of the mechanisms that underlie time-varying interactions in physical and biological systems.
Authors: Martin Wistuba, Prabhu Teja Sivaprasad, Lukas Balles, Giovanni Zappella
Abstract: Recent Continual Learning (CL) methods have combined pretrained Transformers with prompt tuning, a parameter-efficient fine-tuning (PEFT) technique. We argue that the choice of prompt tuning in prior works was an undefended and unablated decision, which has been uncritically adopted by subsequent research, but warrants further research to understand its implications. In this paper, we conduct this research and find that the choice of prompt tuning as a PEFT method hurts the overall performance of the CL system. To illustrate this, we replace prompt tuning with LoRA in two state-of-the-art continual learning methods: Learning to Prompt and S-Prompts. These variants consistently achieve higher accuracy across a wide range of domain-incremental and class-incremental benchmarks, while being competitive in inference speed. Our work highlights a crucial argument: unexamined choices can hinder progress in the field, and rigorous ablations, such as the PEFT method, are required to drive meaningful adoption of CL techniques in real-world applications.
Authors: Inwoo Hwang, Yunhyeok Kwak, Suhyung Choi, Byoung-Tak Zhang, Sanghack Lee
Abstract: Causal dynamics learning has recently emerged as a promising approach to enhancing robustness in reinforcement learning (RL). Typically, the goal is to build a dynamics model that makes predictions based on the causal relationships among the entities. Despite the fact that causal connections often manifest only under certain contexts, existing approaches overlook such fine-grained relationships and lack a detailed understanding of the dynamics. In this work, we propose a novel dynamics model that infers fine-grained causal structures and employs them for prediction, leading to improved robustness in RL. The key idea is to jointly learn the dynamics model with a discrete latent variable that quantizes the state-action space into subgroups. This leads to recognizing meaningful context that displays sparse dependencies, where causal structures are learned for each subgroup throughout the training. Experimental results demonstrate the robustness of our method to unseen states and locally spurious correlations in downstream tasks where fine-grained causal reasoning is crucial. We further illustrate the effectiveness of our subgroup-based approach with quantization in discovering fine-grained causal relationships compared to prior methods.
Authors: Hanming Yang, Antonio Khalil Moretti, Sebastian Macaluso, Philippe Chlenski, Christian A. Naesseth, Itsik Pe'er
Abstract: Reconstructing jets, which provide vital insights into the properties and histories of subatomic particles produced in high-energy collisions, is a main problem in data analyses in collider physics. This intricate task deals with estimating the latent structure of a jet (binary tree) and involves parameters such as particle energy, momentum, and types. While Bayesian methods offer a natural approach for handling uncertainty and leveraging prior knowledge, they face significant challenges due to the super-exponential growth of potential jet topologies as the number of observed particles increases. To address this, we introduce a Combinatorial Sequential Monte Carlo approach for inferring jet latent structures. As a second contribution, we leverage the resulting estimator to develop a variational inference algorithm for parameter learning. Building on this, we introduce a variational family using a pseudo-marginal framework for a fully Bayesian treatment of all variables, unifying the generative model with the inference process. We illustrate our method's effectiveness through experiments using data generated with a collider physics generative model, highlighting superior speed and accuracy across a range of tasks.
Authors: Jiali Nie, Yuanhao Cui, Zhaohui Yang, Weijie Yuan, Xiaojun Jing
Abstract: Extremely Large-scale Array (ELAA) is considered a frontier technology for future communication systems, pivotal in improving wireless systems' rate and spectral efficiency. However, as ELAA employs a multitude of antennas operating at higher frequencies, users are typically situated in the near-field region where the spherical wavefront propagates. This inevitably leads to a significant increase in the overhead of beam training, requiring complex two-dimensional beam searching in both the angle domain and the distance domain. To address this problem, we propose a near-field beamforming method based on unsupervised deep learning. Our convolutional neural network efficiently extracts complex channel state information features by strategically selecting padding and kernel size. We optimize the beamformers to maximize achievable rates in a multi-user network without relying on predefined custom codebooks. Upon deployment, the model requires solely the input of pre-estimated channel state information to derive the optimal beamforming vector. Simulation results show that our proposed scheme can obtain stable beamforming gain compared with the baseline scheme. Furthermore, owing to the inherent traits of deep learning methodologies, this approach substantially diminishes the beam training costs in near-field regions.
Authors: Akshit Sinha, Sreeram Vennam, Charu Sharma, Ponnurangam Kumaraguru
Abstract: Recent advancements in graph learning contributed to explaining predictions generated by Graph Neural Networks. However, existing methodologies often fall short when applied to real-world datasets. We introduce HOGE, a framework to capture higher-order structures using cell complexes, which excel at modeling higher-order relationships. In the real world, higher-order structures are ubiquitous like in molecules or social networks, thus our work significantly enhances the practical applicability of graph explanations. HOGE produces clearer and more accurate explanations compared to prior methods. Our method can be integrated with all existing graph explainers, ensuring seamless integration into current frameworks. We evaluate on GraphXAI benchmark datasets, HOGE achieves improved or comparable performance with minimal computational overhead. Ablation studies show that the performance gain observed can be attributed to the higher-order structures that come from introducing cell complexes.
Authors: Jinqiu Jin, Haoxuan Li, Fuli Feng
Abstract: Fairness has become a crucial aspect in the development of trustworthy machine learning algorithms. Current fairness metrics to measure the violation of demographic parity have the following drawbacks: (i) the average difference of model predictions on two groups cannot reflect their distribution disparity, and (ii) the overall calculation along all possible predictions conceals the extreme local disparity at or around certain predictions. In this work, we propose a novel fairness metric called Maximal Cumulative ratio Disparity along varying Predictions' neighborhood (MCDP), for measuring the maximal local disparity of the fairness-aware classifiers. To accurately and efficiently calculate the MCDP, we develop a provably exact and an approximate calculation algorithm that greatly reduces the computational complexity with low estimation error. We further propose a bi-level optimization algorithm using a differentiable approximation of the MCDP for improving the algorithmic fairness. Extensive experiments on both tabular and image datasets validate that our fair training algorithm can achieve superior fairness-accuracy trade-offs.
Authors: Patryk B\k{e}dkowski, Jan Dubi\'nski, Kamil Deja, Przemys{\l}aw Rokita
Abstract: Simulating detector responses is a crucial part of understanding the inner-workings of particle collisions in the Large Hadron Collider at CERN. The current reliance on statistical Monte-Carlo simulations strains CERN's computational grid, underscoring the urgency for more efficient alternatives. Addressing these challenges, recent proposals advocate for generative machine learning methods. In this study, we present an innovative deep learning simulation approach tailored for the proton Zero Degree Calorimeter in the ALICE experiment. Leveraging a Generative Adversarial Network model with Selective Diversity Increase loss, we directly simulate calorimeter responses. To enhance its capabilities in modeling a broad range of calorimeter response intensities, we expand the SDI-GAN architecture with additional regularization. Moreover, to improve the spatial fidelity of the generated data, we introduce an auxiliary regressor network. Our method offers a significant speedup when comparing to the traditional Monte-Carlo based approaches.
Authors: Mohamed Elsayed, Homayoon Farrahi, Felix Dangel, A. Rupam Mahmood
Abstract: Second-order information is valuable for many applications but challenging to compute. Several works focus on computing or approximating Hessian diagonals, but even this simplification introduces significant additional costs compared to computing a gradient. In the absence of efficient exact computation schemes for Hessian diagonals, we revisit an early approximation scheme proposed by Becker and LeCun (1989, BL89), which has a cost similar to gradients and appears to have been overlooked by the community. We introduce HesScale, an improvement over BL89, which adds negligible extra computation. On small networks, we find that this improvement is of higher quality than all alternatives, even those with theoretical guarantees, such as unbiasedness, while being much cheaper to compute. We use this insight in reinforcement learning problems where small networks are used and demonstrate HesScale in second-order optimization and scaling the step-size parameter. In our experiments, HesScale optimizes faster than existing methods and improves stability through step-size scaling. These findings are promising for scaling second-order methods in larger models in the future.
Authors: F\'elix Therrien, Edward H. Sargent, Oleksandr Voznyy
Abstract: Graph neural networks (GNNs) have emerged as powerful tools to accurately predict materials and molecular properties in computational discovery pipelines. In this article, we exploit the invertible nature of these neural networks to directly generate molecular structures with desired electronic properties. Starting from a random graph or an existing molecule, we perform a gradient ascent while holding the GNN weights fixed in order to optimize its input, the molecular graph, towards the target property. Valence rules are enforced strictly through a judicious graph construction. The method relies entirely on the property predictor; no additional training is required on molecular structures. We demonstrate the application of this method by generating molecules with specific DFT-verified energy gaps and octanol-water partition coefficients (logP). Our approach hits target properties with rates comparable to or better than state-of-the-art generative models while consistently generating more diverse molecules.
Authors: Anke Tang, Li Shen, Yong Luo, Han Hu, Bo Do, Dacheng Tao
Abstract: Deep model fusion is an emerging technique that unifies the predictions or parameters of several deep neural networks into a single model in a cost-effective and data-efficient manner. This enables the unified model to take advantage of the original models' strengths, potentially exceeding their performance. Although a variety of deep model fusion techniques have been introduced, their evaluations tend to be inconsistent and often inadequate to validate their effectiveness and robustness against distribution shifts. To address this issue, we introduce FusionBench, which is the first comprehensive benchmark dedicated to deep model fusion. FusionBench covers a wide range of tasks, including open-vocabulary image classification, text classification, and text-to-text generation. Each category includes up to eight tasks with corresponding task-specific models, featuring both full fine-tuning and LoRA fine-tuning, as well as models of different sizes, to ensure fair and balanced comparisons of various multi-task model fusion techniques across different tasks, model scales, and fine-tuning strategies. We implement and evaluate a broad spectrum of deep model fusion techniques. These techniques range from model ensemble methods, which combine the predictions to improve the overall performance, to model merging, which integrates different models into a single one, and model mixing methods, which upscale or recombine the components of the original models. FusionBench now contains 26 distinct tasks, 74 fine-tuned models, and 16 fusion techniques, and we are committed to consistently expanding the benchmark with more tasks, models, and fusion techniques. In addition, we offer a well-documented set of resources and guidelines to aid researchers in understanding and replicating the benchmark results. Homepage https://tanganke.github.io/fusion_bench/
Authors: Tiago da Silva, Luiz Max Carvalho, Amauri Souza, Samuel Kaski, Diego Mesquita
Abstract: GFlowNets are a promising alternative to MCMC sampling for discrete compositional random variables. Training GFlowNets requires repeated evaluations of the unnormalized target distribution or reward function. However, for large-scale posterior sampling, this may be prohibitive since it incurs traversing the data several times. Moreover, if the data are distributed across clients, employing standard GFlowNets leads to intensive client-server communication. To alleviate both these issues, we propose embarrassingly parallel GFlowNet (EP-GFlowNet). EP-GFlowNet is a provably correct divide-and-conquer method to sample from product distributions of the form $R(\cdot) \propto R_1(\cdot) ... R_N(\cdot)$ -- e.g., in parallel or federated Bayes, where each $R_n$ is a local posterior defined on a data partition. First, in parallel, we train a local GFlowNet targeting each $R_n$ and send the resulting models to the server. Then, the server learns a global GFlowNet by enforcing our newly proposed \emph{aggregating balance} condition, requiring a single communication step. Importantly, EP-GFlowNets can also be applied to multi-objective optimization and model reuse. Our experiments illustrate the EP-GFlowNets's effectiveness on many tasks, including parallel Bayesian phylogenetics, multi-objective multiset, sequence generation, and federated Bayesian structure learning.
Authors: Gijs de Jong, Macha J. Meijer, Derck W. E. Prinzhorn, Harold Ruiter
Abstract: This work aims to reproduce the findings of the paper "Fair Attribute Completion on Graph with Missing Attributes" written by Guo, Chu, and Li arXiv:2302.12977 by investigating the claims made in the paper. This paper suggests that the results of the original paper are reproducible and thus, the claims hold. However, the claim that FairAC is a generic framework for many downstream tasks is very broad and could therefore only be partially tested. Moreover, we show that FairAC is generalizable to various datasets and sensitive attributes and show evidence that the improvement in group fairness of the FairAC framework does not come at the expense of individual fairness. Lastly, the codebase of FairAC has been refactored and is now easily applicable for various datasets and models.
Authors: Yu Zhang, Rui Yu, Zhipeng Yao, Wenyuan Zhang, Jun Wang, Liming Zhang
Abstract: The Mean Square Error (MSE) is commonly utilized to estimate the solution of the optimal value function in the vast majority of offline reinforcement learning (RL) models and has achieved outstanding performance. However, we find that its principle can lead to overestimation phenomenon for the value function. In this paper, we first theoretically analyze overestimation phenomenon led by MSE and provide the theoretical upper bound of the overestimated error. Furthermore, to address it, we propose a novel Bellman underestimated operator to counteract overestimation phenomenon and then prove its contraction characteristics. At last, we propose the offline RL algorithm based on underestimated operator and diffusion policy model. Extensive experimental results on D4RL tasks show that our method can outperform state-of-the-art offline RL algorithms, which demonstrates that our theoretical analysis and underestimation way are effective for offline RL tasks.
Authors: Hrittik Roy, Marco Miani, Carl Henrik Ek, Philipp Hennig, Marvin Pf\"ortner, Lukas Tatzel, S{\o}ren Hauberg
Abstract: Current approximate posteriors in Bayesian neural networks (BNNs) exhibit a crucial limitation: they fail to maintain invariance under reparameterization, i.e. BNNs assign different posterior densities to different parametrizations of identical functions. This creates a fundamental flaw in the application of Bayesian principles as it breaks the correspondence between uncertainty over the parameters with uncertainty over the parametrized function. In this paper, we investigate this issue in the context of the increasingly popular linearized Laplace approximation. Specifically, it has been observed that linearized predictives alleviate the common underfitting problems of the Laplace approximation. We develop a new geometric view of reparametrizations from which we explain the success of linearization. Moreover, we demonstrate that these reparameterization invariance properties can be extended to the original neural network predictive using a Riemannian diffusion process giving a straightforward algorithm for approximate posterior sampling, which empirically improves posterior fit.
Authors: \c{C}a\u{g}lar H{\i}zl{\i}, \c{C}a\u{g}atay Y{\i}ld{\i}z, Matthias Bethge, ST John, Pekka Marttinen
Abstract: This work aims to improve generalization and interpretability of dynamical systems by recovering the underlying lower-dimensional latent states and their time evolutions. Previous work on disentangled representation learning within the realm of dynamical systems focused on the latent states, possibly with linear transition approximations. As such, they cannot identify nonlinear transition dynamics, and hence fail to reliably predict complex future behavior. Inspired by the advances in nonlinear ICA, we propose a state-space modeling framework in which we can identify not just the latent states but also the unknown transition function that maps the past states to the present. We introduce a practical algorithm based on variational auto-encoders and empirically demonstrate in realistic synthetic settings that we can (i) recover latent state dynamics with high accuracy, (ii) correspondingly achieve high future prediction accuracy, and (iii) adapt fast to new environments.
Authors: Hiroaki Chiba-Okabe, Weijie J. Su
Abstract: The rapid progress of generative AI technology has sparked significant copyright concerns, leading to numerous lawsuits filed against AI developers. While some studies explore methods to mitigate copyright risks by steering the outputs of generative models away from those resembling copyrighted data, little attention has been paid to the question of how much of a resemblance is undesirable; more original or unique data are afforded stronger protection, and the threshold level of resemblance for constituting infringement correspondingly lower. Here, leveraging this principle, we propose a genericization method that modifies the outputs of a generative model to make them more generic and less likely to infringe copyright. To achieve this, we introduce a metric for quantifying the level of originality of data in a manner that is consistent with the legal framework. This metric can be practically estimated by drawing samples from a generative model, which is then used for the genericization process. Experiments demonstrate that our genericization method successfully modifies the output of a text-to-image generative model so that it produces more generic, copyright-compliant images.
Authors: Tianren Zhang, Chujie Zhao, Guanyu Chen, Yizhou Jiang, Feng Chen
Abstract: Learning representations that generalize under distribution shifts is critical for building robust machine learning models. However, despite significant efforts in recent years, algorithmic advances in this direction have been limited. In this work, we seek to understand the fundamental difficulty of out-of-distribution generalization with deep neural networks. We first empirically show that perhaps surprisingly, even allowing a neural network to explicitly fit the representations obtained from a teacher network that can generalize out-of-distribution is insufficient for the generalization of the student network. Then, by a theoretical study of two-layer ReLU networks optimized by stochastic gradient descent (SGD) under a structured feature model, we identify a fundamental yet unexplored feature learning proclivity of neural networks, feature contamination: neural networks can learn uncorrelated features together with predictive features, resulting in generalization failure under distribution shifts. Notably, this mechanism essentially differs from the prevailing narrative in the literature that attributes the generalization failure to spurious correlations. Overall, our results offer new insights into the non-linear feature learning dynamics of neural networks and highlight the necessity of considering inductive biases in out-of-distribution generalization.
Authors: Nicolo Colombo
Abstract: Conformal Prediction (CP) algorithms estimate the uncertainty of a prediction model by calibrating its outputs on labeled data. The same calibration scheme usually applies to any model and data without modifications. The obtained prediction intervals are valid by construction but could be inefficient, i.e. unnecessarily big, if the prediction errors are not uniformly distributed over the input space. We present a general scheme to localize the intervals by training the calibration process. The standard prediction error is replaced by an optimized distance metric that depends explicitly on the object attributes. Learning the optimal metric is equivalent to training a Normalizing Flow that acts on the joint distribution of the errors and the inputs. Unlike the Error Re-weighting CP algorithm of Papadopoulos et al. (2008), the framework allows estimating the gap between nominal and empirical conditional validity. The approach is compatible with existing locally-adaptive CP strategies based on re-weighting the calibration samples and applies to any point-prediction model without retraining.
Authors: Marius Lindauer, Florian Karl, Anne Klier, Julia Moosbauer, Alexander Tornede, Andreas Mueller, Frank Hutter, Matthias Feurer, Bernd Bischl
Abstract: Automated machine learning (AutoML) was formed around the fundamental objectives of automatically and efficiently configuring machine learning (ML) workflows, aiding the research of new ML algorithms, and contributing to the democratization of ML by making it accessible to a broader audience. Over the past decade, commendable achievements in AutoML have primarily focused on optimizing predictive performance. This focused progress, while substantial, raises questions about how well AutoML has met its broader, original goals. In this position paper, we argue that a key to unlocking AutoML's full potential lies in addressing the currently underexplored aspect of user interaction with AutoML systems, including their diverse roles, expectations, and expertise. We envision a more human-centered approach in future AutoML research, promoting the collaborative design of ML systems that tightly integrates the complementary strengths of human expertise and AutoML methodologies.
Authors: Tanguy Lefort, Antoine Affouard, Benjamin Charlier, Jean-Christophe Lombardo, Mathias Chouet, Herv\'e Go\"eau, Joseph Salmon, Pierre Bonnet, Alexis Joly
Abstract: Deep learning models for plant species identification rely on large annotated datasets. The PlantNet system enables global data collection by allowing users to upload and annotate plant observations, leading to noisy labels due to diverse user skills. Achieving consensus is crucial for training, but the vast scale of collected data makes traditional label aggregation strategies challenging. Existing methods either retain all observations, resulting in noisy training data or selectively keep those with sufficient votes, discarding valuable information. Additionally, as many species are rarely observed, user expertise can not be evaluated as an inter-user agreement: otherwise, botanical experts would have a lower weight in the AI training step than the average user. Our proposed label aggregation strategy aims to cooperatively train plant identification AI models. This strategy estimates user expertise as a trust score per user based on their ability to identify plant species from crowdsourced data. The trust score is recursively estimated from correctly identified species given the current estimated labels. This interpretable score exploits botanical experts' knowledge and the heterogeneity of users. Subsequently, our strategy removes unreliable observations but retains those with limited trusted annotations, unlike other approaches. We evaluate PlantNet's strategy on a released large subset of the PlantNet database focused on European flora, comprising over 6M observations and 800K users. We demonstrate that estimating users' skills based on the diversity of their expertise enhances labeling performance. Our findings emphasize the synergy of human annotation and data filtering in improving AI performance for a refined dataset. We explore incorporating AI-based votes alongside human input. This can further enhance human-AI interactions to detect unreliable observations.
Authors: Micha{\l} Zawalski, Gracjan G\'oral, Micha{\l} Tyrolski, Emilia Wi\'snios, Franciszek Budrowski, {\L}ukasz Kuci\'nski, Piotr Mi{\l}o\'s
Abstract: Efficiently tackling combinatorial reasoning problems, particularly the notorious NP-hard tasks, remains a significant challenge for AI research. Recent efforts have sought to enhance planning by incorporating hierarchical high-level search strategies, known as subgoal methods. While promising, their performance against traditional low-level planners is inconsistent, raising questions about their application contexts. In this study, we conduct an in-depth exploration of subgoal-planning methods for combinatorial reasoning. We identify the attributes pivotal for leveraging the advantages of high-level search: hard-to-learn value functions, complex action spaces, presence of dead ends in the environment, or using data collected from diverse experts. We propose a consistent evaluation methodology to achieve meaningful comparisons between methods and reevaluate the state-of-the-art algorithms.
Authors: Dexiong Chen, Till Hendrik Schulz, Karsten Borgwardt
Abstract: Message-passing graph neural networks (GNNs), while excelling at capturing local relationships, often struggle with long-range dependencies on graphs. Conversely, graph transformers (GTs) enable information exchange between all nodes but oversimplify the graph structure by treating them as a set of fixed-length vectors. This work proposes a novel architecture, NeuralWalker, that overcomes the limitations of both methods by combining random walks with message passing. NeuralWalker achieves this by treating random walks as sequences, allowing for the application of recent advances in sequence models in order to capture long-range dependencies within these walks. Based on this concept, we propose a framework that offers (1) more expressive graph representations through random walk sequences, (2) the ability to utilize any sequence model for capturing long-range dependencies, and (3) the flexibility by integrating various GNN and GT architectures. Our experimental evaluations demonstrate that NeuralWalker achieves significant performance improvements on 19 graph and node benchmark datasets, notably outperforming existing methods by up to 13% on the PascalVoc-SP and COCO-SP datasets. Code is available at https://github.com/BorgwardtLab/NeuralWalker.
Authors: Pio Calderon, Marian-Andrei Rizoiu
Abstract: The spread of content on social media is shaped by intertwining factors on three levels: the source, the content itself, and the pathways of content spread. At the lowest level, the popularity of the sharing user determines its eventual reach. However, higher-level factors such as the nature of the online item and the credibility of its source also play crucial roles in determining how widely and rapidly the online item spreads. In this work, we propose the Bayesian Mixture Hawkes (BMH) model to jointly learn the influence of source, content and spread. We formulate the BMH model as a hierarchical mixture model of separable Hawkes processes, accommodating different classes of Hawkes dynamics and the influence of feature sets on these classes. We test the BMH model on two learning tasks, cold-start popularity prediction and temporal profile generalization performance, applying to two real-world retweet cascade datasets referencing articles from controversial and traditional media publishers. The BMH model outperforms the state-of-the-art models and predictive baselines on both datasets and utilizes cascade- and item-level information better than the alternatives. Lastly, we perform a counter-factual analysis where we apply the trained publisher-level BMH models to a set of article headlines and show that effectiveness of headline writing style (neutral, clickbait, inflammatory) varies across publishers. The BMH model unveils differences in style effectiveness between controversial and reputable publishers, where we find clickbait to be notably more effective for reputable publishers as opposed to controversial ones, which links to the latter's overuse of clickbait.
Authors: Haozhe Chen, Andres Felipe Duque Correa, Guy Wolf, Kevin R. Moon
Abstract: Data visualization via dimensionality reduction is an important tool in exploratory data analysis. However, when the data are noisy, many existing methods fail to capture the underlying structure of the data. The method called Empirical Intrinsic Geometry (EIG) was previously proposed for performing dimensionality reduction on high dimensional dynamical processes while theoretically eliminating all noise. However, implementing EIG in practice requires the construction of high-dimensional histograms, which suffer from the curse of dimensionality. Here we propose a new data visualization method called Functional Information Geometry (FIG) for dynamical processes that adapts the EIG framework while using approaches from functional data analysis to mitigate the curse of dimensionality. We experimentally demonstrate that the resulting method outperforms a variant of EIG designed for visualization in terms of capturing the true structure, hyperparameter robustness, and computational speed. We then use our method to visualize EEG brain measurements of sleep activity.
Authors: Salim rezvani, Farhad Pourpanah, Chee Peng Lim, Q. M. Jonathan Wu
Abstract: This paper presents a review on methods for class-imbalanced learning with the Support Vector Machine (SVM) and its variants. We first explain the structure of SVM and its variants and discuss their inefficiency in learning with class-imbalanced data sets. We introduce a hierarchical categorization of SVM-based models with respect to class-imbalanced learning. Specifically, we categorize SVM-based models into re-sampling, algorithmic, and fusion methods, and discuss the principles of the representative models in each category. In addition, we conduct a series of empirical evaluations to compare the performances of various representative SVM-based models in each category using benchmark imbalanced data sets, ranging from low to high imbalanced ratios. Our findings reveal that while algorithmic methods are less time-consuming owing to no data pre-processing requirements, fusion methods, which combine both re-sampling and algorithmic approaches, generally perform the best, but with a higher computational load. A discussion on research gaps and future research directions is provided.
Authors: Jinsheng Yuan, Zhuangkun Wei, Weisi Guo
Abstract: Over-the-Air Federated Learning (OTA-FL) has been extensively investigated as a privacy-preserving distributed learning mechanism. Realistic systems will see FL clients with diverse size, weight, and power configurations. A critical research gap in existing OTA-FL research is the assumption of homogeneous client computational bit precision. Indeed, many clients may exploit approximate computing (AxC) where bit precisions are adjusted for energy and computational efficiency. The dynamic distribution of bit precision updates amongst FL clients poses an open challenge for OTA-FL, as is is incompatible in the wireless modulation superposition space. Here, we propose an AxC-based OTA-FL framework of clients with multiple precisions, demonstrating the following innovations: (i) optimize the quantization-performance trade-off for both server and clients within the constraints of varying edge computing capabilities and learning accuracy requirements, and (ii) develop heterogeneous gradient resolution OTA-FL modulation schemes to ensure compatibility with physical layer OTA aggregation. Our findings indicate that we can design modulation schemes that enable AxC based OTA-FL, which can achieve 50\% faster and smoother server convergence and a performance enhancement for the lowest precision clients compared to a homogeneous precision approach. This demonstrates the great potential of our AxC-based OTA-FL approach in heterogeneous edge computing environments.
Authors: Kangyu Zheng, Yingzhou Lu, Zaixi Zhang, Zhongwei Wan, Yao Ma, Marinka Zitnik, Tianfan Fu
Abstract: Currently, the field of structure-based drug design is dominated by three main types of algorithms: search-based algorithms, deep generative models, and reinforcement learning. While existing works have typically focused on comparing models within a single algorithmic category, cross-algorithm comparisons remain scarce. In this paper, to fill the gap, we establish a benchmark to evaluate the performance of sixteen models across these different algorithmic foundations by assessing the pharmaceutical properties of the generated molecules and their docking affinities with specified target proteins. We highlight the unique advantages of each algorithmic approach and offer recommendations for the design of future SBDD models. We emphasize that 1D/2D ligand-centric drug design methods can be used in SBDD by treating the docking function as a black-box oracle, which is typically neglected. The empirical results show that 1D/2D methods achieve competitive performance compared with 3D-based methods that use the 3D structure of the target protein explicitly. Also, AutoGrow4, a 2D molecular graph-based genetic algorithm, dominates SBDD in terms of optimization ability. The relevant code is available in https://github.com/zkysfls/2024-sbdd-benchmark.
Authors: Wei Shao, Rongyi Zhu, Cai Yang, Chandra Thapa, Muhammad Ejaz Ahmed, Seyit Camtepe, Rui Zhang, DuYong Kim, Hamid Menouar, Flora D. Salim
Abstract: Spatiotemporal data is prevalent in a wide range of edge devices, such as those used in personal communication and financial transactions. Recent advancements have sparked a growing interest in integrating spatiotemporal analysis with large-scale language models. However, spatiotemporal data often contains sensitive information, making it unsuitable for open third-party access. To address this challenge, we propose a Graph-GAN-based model for generating privacy-protected spatiotemporal data. Our approach incorporates spatial and temporal attention blocks in the discriminator and a spatiotemporal deconvolution structure in the generator. These enhancements enable efficient training under Gaussian noise to achieve differential privacy. Extensive experiments conducted on three real-world spatiotemporal datasets validate the efficacy of our model. Our method provides a privacy guarantee while maintaining the data utility. The prediction model trained on our generated data maintains a competitive performance compared to the model trained on the original data.
Authors: Sifat Ut Taki, Spyridon Mastorakis
Abstract: Training a proprietary Neural Network (NN) model with a proprietary dataset on the cloud comes at the risk of exposing the model architecture and the dataset to the cloud service provider. To tackle this problem, in this paper, we present an NN obfuscation framework, called Amalgam, to train NN models in a privacy-preserving manner in existing cloud-based environments. Amalgam achieves that by augmenting NN models and the datasets to be used for training with well-calibrated noise to "hide" both the original model architectures and training datasets from the cloud. After training, Amalgam extracts the original models from the augmented models and returns them to users. Our evaluation results with different computer vision and natural language processing models and datasets demonstrate that Amalgam: (i) introduces modest overheads into the training process without impacting its correctness, and (ii) does not affect the model's accuracy.
Authors: Wen-Yu Xi, Juan Wang, Yu-Lin Zhang, Jin-Xing Liu, Yin-Lian Gao
Abstract: The emerging research shows that lncRNA has crucial research value in a series of complex human diseases. Therefore, the accurate identification of lncRNA-disease associations (LDAs) is very important for the warning and treatment of diseases. However, most of the existing methods have limitations in identifying nonlinear LDAs, and it remains a huge challenge to predict new LDAs. In this paper, a deep learning model based on a heterogeneous network and convolutional neural network (CNN) is proposed for lncRNA-disease association prediction, named HCNNLDA. The heterogeneous network containing the lncRNA, disease, and miRNA nodes, is constructed firstly. The embedding matrix of a lncRNA-disease node pair is constructed according to various biological premises about lncRNAs, diseases, and miRNAs. Then, the low-dimensional feature representation is fully learned by the convolutional neural network. In the end, the XGBoot classifier model is trained to predict the potential LDAs. HCNNLDA obtains a high AUC value of 0.9752 and AUPR of 0.9740 under the 5-fold cross-validation. The experimental results show that the proposed model has better performance than that of several latest prediction models. Meanwhile, the effectiveness of HCNNLDA in identifying novel LDAs is further demonstrated by case studies of three diseases. To sum up, HCNNLDA is a feasible calculation model to predict LDAs.
Authors: Siddharth Nair, Timothy F. Walsh, Greg Pickrell, Fabio Semperlotti
Abstract: In this paper, we introduce a physics and geometry informed neural operator network with application to the forward simulation of acoustic scattering. The development of geometry informed deep learning models capable of learning a solution operator for different computational domains is a problem of general importance for a variety of engineering applications. To this end, we propose a physics-informed deep operator network (DeepONet) capable of predicting the scattered pressure field for arbitrarily shaped scatterers using a geometric parameterization approach based on non-uniform rational B-splines (NURBS). This approach also results in parsimonious representations of non-trivial scatterer geometries. In contrast to existing physics-based approaches that require model re-evaluation when changing the computational domains, our trained model is capable of learning solution operator that can approximate physically-consistent scattered pressure field in just a few seconds for arbitrary rigid scatterer shapes; it follows that the computational time for forward simulations can improve (i.e. be reduced) by orders of magnitude in comparison to the traditional forward solvers. In addition, this approach can evaluate the scattered pressure field without the need for labeled training data. After presenting the theoretical approach, a comprehensive numerical study is also provided to illustrate the remarkable ability of this approach to simulate the acoustic pressure fields resulting from arbitrary combinations of arbitrary scatterer geometries. These results highlight the unique generalization capability of the proposed operator learning approach.
Authors: Jinyin Chen, Xiaoming Zhao, Haibin Zheng, Xiao Li, Sheng Xiang, Haifeng Guo
Abstract: Benefiting from well-trained deep neural networks (DNNs), model compression have captured special attention for computing resource limited equipment, especially edge devices. Knowledge distillation (KD) is one of the widely used compression techniques for edge deployment, by obtaining a lightweight student model from a well-trained teacher model released on public platforms. However, it has been empirically noticed that the backdoor in the teacher model will be transferred to the student model during the process of KD. Although numerous KD methods have been proposed, most of them focus on the distillation of a high-performing student model without robustness consideration. Besides, some research adopts KD techniques as effective backdoor mitigation tools, but they fail to perform model compression at the same time. Consequently, it is still an open problem to well achieve two objectives of robust KD, i.e., student model's performance and backdoor mitigation. To address these issues, we propose RobustKD, a robust knowledge distillation that compresses the model while mitigating backdoor based on feature variance. Specifically, RobustKD distinguishes the previous works in three key aspects: (1) effectiveness: by distilling the feature map of the teacher model after detoxification, the main task performance of the student model is comparable to that of the teacher model; (2) robustness: by reducing the characteristic variance between the teacher model and the student model, it mitigates the backdoor of the student model under backdoored teacher model scenario; (3) generic: RobustKD still has good performance in the face of multiple data models (e.g., WRN 28-4, Pyramid-200) and diverse DNNs (e.g., ResNet50, MobileNet).
Authors: Tim Franzmeyer, Aleksandar Shtedritski, Samuel Albanie, Philip Torr, Jo\~ao F. Henriques, Jakob N. Foerster
Abstract: Benchmarks have been essential for driving progress in machine learning. A better understanding of LLM capabilities on real world tasks is vital for safe development. Designing adequate LLM benchmarks is challenging: Data from real-world tasks is hard to collect, public availability of static evaluation data results in test data contamination and benchmark overfitting, and periodically generating new evaluation data is tedious and may result in temporally inconsistent results. We introduce HelloFresh, based on continuous streams of real-world data generated by intrinsically motivated human labelers. It covers recent events from X (formerly Twitter) community notes and edits of Wikipedia pages, mitigating the risk of test data contamination and benchmark overfitting. Any X user can propose an X note to add additional context to a misleading post (formerly tweet); if the community classifies it as helpful, it is shown with the post. Similarly, Wikipedia relies on community-based consensus, allowing users to edit articles or revert edits made by other users. Verifying whether an X note is helpful or whether a Wikipedia edit should be accepted are hard tasks that require grounding by querying the web. We backtest state-of-the-art LLMs supplemented with simple web search access and find that HelloFresh yields a temporally consistent ranking. To enable continuous evaluation on HelloFresh, we host a public leaderboard and periodically updated evaluation data at https://tinyurl.com/hello-fresh-LLM.
Authors: Imad Aouali, Victor-Emmanuel Brunel, David Rohde, Anna Korba
Abstract: Off-policy learning (OPL) often involves minimizing a risk estimator based on importance weighting to correct bias from the logging policy used to collect data. However, this method can produce an estimator with a high variance. A common solution is to regularize the importance weights and learn the policy by minimizing an estimator with penalties derived from generalization bounds specific to the estimator. This approach, known as pessimism, has gained recent attention but lacks a unified framework for analysis. To address this gap, we introduce a comprehensive PAC-Bayesian framework to examine pessimism with regularized importance weighting. We derive a tractable PAC-Bayesian generalization bound that universally applies to common importance weight regularizations, enabling their comparison within a single framework. Our empirical results challenge common understanding, demonstrating the effectiveness of standard IW regularization techniques.
Authors: Akhil Jalan, Arya Mazumdar, Soumendu Sundar Mukherjee, Purnamrita Sarkar
Abstract: We study transfer learning for estimation in latent variable network models. In our setting, the conditional edge probability matrices given the latent variables are represented by $P$ for the source and $Q$ for the target. We wish to estimate $Q$ given two kinds of data: (1) edge data from a subgraph induced by an $o(1)$ fraction of the nodes of $Q$, and (2) edge data from all of $P$. If the source $P$ has no relation to the target $Q$, the estimation error must be $\Omega(1)$. However, we show that if the latent variables are shared, then vanishing error is possible. We give an efficient algorithm that utilizes the ordering of a suitably defined graph distance. Our algorithm achieves $o(1)$ error and does not assume a parametric form on the source or target networks. Next, for the specific case of Stochastic Block Models we prove a minimax lower bound and show that a simple algorithm achieves this rate. Finally, we empirically demonstrate our algorithm's use on real-world and simulated graph transfer problems.
Authors: Tianyi Zhou, Deqing Fu, Vatsal Sharan, Robin Jia
Abstract: Pre-trained large language models (LLMs) exhibit impressive mathematical reasoning capabilities, yet how they compute basic arithmetic, such as addition, remains unclear. This paper shows that pre-trained LLMs add numbers using Fourier features -- dimensions in the hidden state that represent numbers via a set of features sparse in the frequency domain. Within the model, MLP and attention layers use Fourier features in complementary ways: MLP layers primarily approximate the magnitude of the answer using low-frequency features, while attention layers primarily perform modular addition (e.g., computing whether the answer is even or odd) using high-frequency features. Pre-training is crucial for this mechanism: models trained from scratch to add numbers only exploit low-frequency features, leading to lower accuracy. Introducing pre-trained token embeddings to a randomly initialized model rescues its performance. Overall, our analysis demonstrates that appropriate pre-trained representations (e.g., Fourier features) can unlock the ability of Transformers to learn precise mechanisms for algorithmic tasks.
Authors: Saba Ahmadi, Siddharth Bhandari, Avrim Blum, Chen Dan, Prabhav Jain
Abstract: A major challenge in defending against adversarial attacks is the enormous space of possible attacks that even a simple adversary might perform. To address this, prior work has proposed a variety of defenses that effectively reduce the size of this space. These include randomized smoothing methods that add noise to the input to take away some of the adversary's impact. Another approach is input discretization which limits the adversary's possible number of actions. Motivated by these two approaches, we introduce a new notion of adversarial loss which we call distributional adversarial loss, to unify these two forms of effectively weakening an adversary. In this notion, we assume for each original example, the allowed adversarial perturbation set is a family of distributions (e.g., induced by a smoothing procedure), and the adversarial loss over each example is the maximum loss over all the associated distributions. The goal is to minimize the overall adversarial loss. We show generalization guarantees for our notion of adversarial loss in terms of the VC-dimension of the hypothesis class and the size of the set of allowed adversarial distributions associated with each input. We also investigate the role of randomness in achieving robustness against adversarial attacks in the methods described above. We show a general derandomization technique that preserves the extent of a randomized classifier's robustness against adversarial attacks. We corroborate the procedure experimentally via derandomizing the Random Projection Filters framework of \cite{dong2023adversarial}. Our procedure also improves the robustness of the model against various adversarial attacks.
Authors: Haoyu Han, Juanhui Li, Wei Huang, Xianfeng Tang, Hanqing Lu, Chen Luo, Hui Liu, Jiliang Tang
Abstract: Graph Neural Networks (GNNs) have proven to be highly effective for node classification tasks across diverse graph structural patterns. Traditionally, GNNs employ a uniform global filter, typically a low-pass filter for homophilic graphs and a high-pass filter for heterophilic graphs. However, real-world graphs often exhibit a complex mix of homophilic and heterophilic patterns, rendering a single global filter approach suboptimal. In this work, we theoretically demonstrate that a global filter optimized for one pattern can adversely affect performance on nodes with differing patterns. To address this, we introduce a novel GNN framework Node-MoE that utilizes a mixture of experts to adaptively select the appropriate filters for different nodes. Extensive experiments demonstrate the effectiveness of Node-MoE on both homophilic and heterophilic graphs.
Authors: Bruno Machado Pacheco, Eduardo Camponogara
Abstract: This paper introduces Physics-Informed Deep Equilibrium Models (PIDEQs) for solving initial value problems (IVPs) of ordinary differential equations (ODEs). Leveraging recent advancements in deep equilibrium models (DEQs) and physics-informed neural networks (PINNs), PIDEQs combine the implicit output representation of DEQs with physics-informed training techniques. We validate PIDEQs using the Van der Pol oscillator as a benchmark problem, demonstrating their efficiency and effectiveness in solving IVPs. Our analysis includes key hyperparameter considerations for optimizing PIDEQ performance. By bridging deep learning and physics-based modeling, this work advances computational techniques for solving IVPs, with implications for scientific computing and engineering applications.
Authors: Cody Blakeney, Mansheej Paul, Brett W. Larsen, Sean Owen, Jonathan Frankle
Abstract: Pretraining datasets for large language models (LLMs) have grown to trillions of tokens composed of large amounts of CommonCrawl (CC) web scrape along with smaller, domain-specific datasets. It is expensive to understand the impact of these domain-specific datasets on model capabilities as training at large FLOP scales is required to reveal significant changes to difficult and emergent benchmarks. Given the increasing cost of experimenting with pretraining data, how does one determine the optimal balance between the diversity in general web scrapes and the information density of domain specific data? In this work, we show how to leverage the smaller domain specific datasets by upsampling them relative to CC at the end of training to drive performance improvements on difficult benchmarks. This simple technique allows us to improve up to 6.90 pp on MMLU, 8.26 pp on GSM8K, and 6.17 pp on HumanEval relative to the base data mix for a 7B model trained for 1 trillion (T) tokens, thus rivaling Llama-2 (7B)$\unicode{x2014}$a model trained for twice as long. We experiment with ablating the duration of domain upsampling from 5% to 30% of training and find that 10% to 20% percent is optimal for navigating the tradeoff between general language modeling capabilities and targeted benchmarks. We also use domain upsampling to characterize at scale the utility of individual datasets for improving various benchmarks by removing them during this final phase of training. This tool opens up the ability to experiment with the impact of different pretraining datasets at scale, but at an order of magnitude lower cost compared to full pretraining runs.
Authors: Amir Zandieh, Majid Daliri, Insu Han
Abstract: Serving LLMs requires substantial memory due to the storage requirements of Key-Value (KV) embeddings in the KV cache, which grows with sequence length. An effective approach to compress KV cache is quantization. However, traditional quantization methods face significant memory overhead due to the need to store quantization constants (at least a zero point and a scale) in full precision per data block. Depending on the block size, this overhead can add 1 or 2 bits per quantized number. We introduce QJL, a new quantization approach that consists of a Johnson-Lindenstrauss (JL) transform followed by sign-bit quantization. In contrast to existing methods, QJL eliminates memory overheads by removing the need for storing quantization constants. We propose an asymmetric estimator for the inner product of two vectors and demonstrate that applying QJL to one vector and a standard JL transform without quantization to the other provides an unbiased estimator with minimal distortion. We have developed an efficient implementation of the QJL sketch and its corresponding inner product estimator, incorporating a lightweight CUDA kernel for optimized computation. When applied across various LLMs and NLP tasks to quantize the KV cache to only 3 bits, QJL demonstrates a more than fivefold reduction in KV cache memory usage without compromising accuracy, all while achieving faster runtime. Codes are available at \url{https://github.com/amirzandieh/QJL}.
Authors: Yuhui Wang, Weida Li, Francesco Faccio, Qingyuan Wu, J\"urgen Schmidhuber
Abstract: Value iteration networks (VINs) enable end-to-end learning for planning tasks by employing a differentiable "planning module" that approximates the value iteration algorithm. However, long-term planning remains a challenge because training very deep VINs is difficult. To address this problem, we embed highway value iteration -- a recent algorithm designed to facilitate long-term credit assignment -- into the structure of VINs. This improvement augments the "planning module" of the VIN with three additional components: 1) an "aggregate gate," which constructs skip connections to improve information flow across many layers; 2) an "exploration module," crafted to increase the diversity of information and gradient flow in spatial dimensions; 3) a "filter gate" designed to ensure safe exploration. The resulting novel highway VIN can be trained effectively with hundreds of layers using standard backpropagation. In long-term planning tasks requiring hundreds of planning steps, deep highway VINs outperform both traditional VINs and several advanced, very deep NNs.
Authors: Hong Chul Nam, Julius Berner, Anima Anandkumar
Abstract: We propose Neural Walk-on-Spheres (NWoS), a novel neural PDE solver for the efficient solution of high-dimensional Poisson equations. Leveraging stochastic representations and Walk-on-Spheres methods, we develop novel losses for neural networks based on the recursive solution of Poisson equations on spheres inside the domain. The resulting method is highly parallelizable and does not require spatial gradients for the loss. We provide a comprehensive comparison against competing methods based on PINNs, the Deep Ritz method, and (backward) stochastic differential equations. In several challenging, high-dimensional numerical examples, we demonstrate the superiority of NWoS in accuracy, speed, and computational costs. Compared to commonly used PINNs, our approach can reduce memory usage and errors by orders of magnitude. Furthermore, we apply NWoS to problems in PDE-constrained optimization and molecular dynamics to show its efficiency in practical applications.
Authors: Darshil Doshi, Tianyu He, Aritra Das, Andrey Gromov
Abstract: Neural networks readily learn a subset of the modular arithmetic tasks, while failing to generalize on the rest. This limitation remains unmoved by the choice of architecture and training strategies. On the other hand, an analytical solution for the weights of Multi-layer Perceptron (MLP) networks that generalize on the modular addition task is known in the literature. In this work, we (i) extend the class of analytical solutions to include modular multiplication as well as modular addition with many terms. Additionally, we show that real networks trained on these datasets learn similar solutions upon generalization (grokking). (ii) We combine these "expert" solutions to construct networks that generalize on arbitrary modular polynomials. (iii) We hypothesize a classification of modular polynomials into learnable and non-learnable via neural networks training; and provide experimental evidence supporting our claims.
Authors: Konstantinos Roditakis, Spyridon Thermos, Nikolaos Zioulis
Abstract: Despite the recent advances in the so-called "cold start" generation from text prompts, their needs in data and computing resources, as well as the ambiguities around intellectual property and privacy concerns pose certain counterarguments for their utility. An interesting and relatively unexplored alternative has been the introduction of unconditional synthesis from a single sample, which has led to interesting generative applications. In this paper we focus on single-shot motion generation and more specifically on accelerating the training time of a Generative Adversarial Network (GAN). In particular, we tackle the challenge of GAN's equilibrium collapse when using mini-batch training by carefully annealing the weights of the loss functions that prevent mode collapse. Additionally, we perform statistical analysis in the generator and discriminator models to identify correlations between training stages and enable transfer learning. Our improved GAN achieves competitive quality and diversity on the Mixamo benchmark when compared to the original GAN architecture and a single-shot diffusion model, while being up to x6.8 faster in training time from the former and x1.75 from the latter. Finally, we demonstrate the ability of our improved GAN to mix and compose motion with a single forward pass. Project page available at https://moverseai.github.io/single-shot.
Authors: Shijian Deng, Erin E. Kosloski, Siddhi Patel, Zeke A. Barnett, Yiyang Nan, Alexander Kaplan, Sisira Aarukapalli, William T. Doan, Matthew Wang, Harsh Singh, Pamela R. Rollins, Yapeng Tian
Abstract: In this article, we introduce a novel problem of audio-visual autism behavior recognition, which includes social behavior recognition, an essential aspect previously omitted in AI-assisted autism screening research. We define the task at hand as one that is audio-visual autism behavior recognition, which uses audio and visual cues, including any speech present in the audio, to recognize autism-related behaviors. To facilitate this new research direction, we collected an audio-visual autism spectrum dataset (AV-ASD), currently the largest video dataset for autism screening using a behavioral approach. It covers an extensive range of autism-associated behaviors, including those related to social communication and interaction. To pave the way for further research on this new problem, we intensively explored leveraging foundation models and multimodal large language models across different modalities. Our experiments on the AV-ASD dataset demonstrate that integrating audio, visual, and speech modalities significantly enhances the performance in autism behavior recognition. Additionally, we explored the use of a post-hoc to ad-hoc pipeline in a multimodal large language model to investigate its potential to augment the model's explanatory capability during autism behavior recognition. We will release our dataset, code, and pre-trained models.
Authors: Ruizhe Huang, Xiaohui Zhang, Zhaoheng Ni, Li Sun, Moto Hira, Jeff Hwang, Vimal Manohar, Vineel Pratap, Matthew Wiesner, Shinji Watanabe, Daniel Povey, Sanjeev Khudanpur
Abstract: Connectionist temporal classification (CTC) models are known to have peaky output distributions. Such behavior is not a problem for automatic speech recognition (ASR), but it can cause inaccurate forced alignments (FA), especially at finer granularity, e.g., phoneme level. This paper aims at alleviating the peaky behavior for CTC and improve its suitability for forced alignment generation, by leveraging label priors, so that the scores of alignment paths containing fewer blanks are boosted and maximized during training. As a result, our CTC model produces less peaky posteriors and is able to more accurately predict the offset of the tokens besides their onset. It outperforms the standard CTC model and a heuristics-based approach for obtaining CTC's token offset timestamps by 12-40% in phoneme and word boundary errors (PBE and WBE) measured on the Buckeye and TIMIT data. Compared with the most widely used FA toolkit Montreal Forced Aligner (MFA), our method performs similarly on PBE/WBE on Buckeye, yet falls behind MFA on TIMIT. Nevertheless, our method has a much simpler training pipeline and better runtime efficiency. Our training recipe and pretrained model are released in TorchAudio.
Authors: Ognjen Kundacina, Vladimir Vincan, Dragisa Miskovic
Abstract: Emphasizing a data-centric AI approach, this paper introduces a novel two-stage active learning (AL) pipeline for automatic speech recognition (ASR), combining unsupervised and supervised AL methods. The first stage utilizes unsupervised AL by using x-vectors clustering for diverse sample selection from unlabeled speech data, thus establishing a robust initial dataset for the subsequent supervised AL. The second stage incorporates a supervised AL strategy, with a batch AL method specifically developed for ASR, aimed at selecting diverse and informative batches of samples. Here, sample diversity is also achieved using x-vectors clustering, while the most informative samples are identified using a Bayesian AL method tailored for ASR with an adaptation of Monte Carlo dropout to approximate Bayesian inference. This approach enables precise uncertainty estimation, thereby enhancing ASR model training with significantly reduced data requirements. Our method has shown superior performance compared to competing methods on homogeneous, heterogeneous, and OOD test sets, demonstrating that strategic sample selection and innovative Bayesian modeling can substantially optimize both labeling effort and data utilization in deep learning-based ASR applications.
Authors: Shakeel Ahmad Sheikh
Abstract: Speech production is a complex phenomenon, wherein the brain orchestrates a sequence of processes involving thought processing, motor planning, and the execution of articulatory movements. However, this intricate execution of various processes is susceptible to influence and disruption by various neurodegenerative pathological speech disorders, such as Parkinsons' disease, resulting in dysarthria, apraxia, and other conditions. These disorders lead to pathological speech characterized by abnormal speech patterns and imprecise articulation. Diagnosing these speech disorders in clinical settings typically involves auditory perceptual tests, which are time-consuming, and the diagnosis can vary among clinicians based on their experiences, biases, and cognitive load during the diagnosis. Additionally, unlike neurotypical speakers, patients with speech pathologies or impairments are unable to access various virtual assistants such as Alexa, Siri, etc. To address these challenges, several automatic pathological speech detection (PSD) approaches have been proposed. These approaches aim to provide efficient and accurate detection of speech disorders, thereby facilitating timely intervention and support for individuals affected by these conditions. These approaches mainly vary in two aspects: the input representations utilized and the classifiers employed. Due to the limited availability of data, the performance of detection remains subpar. Self-supervised learning (SSL) embeddings, such as wav2vec2, and their multilingual versions, are being explored as a promising avenue to improve performance. These embeddings leverage self-supervised learning techniques to extract rich representations from audio data, thereby offering a potential solution to address the limitations posed by the scarcity of labeled data.
Authors: Trishna Chakraborty, Erfan Shayegani, Zikui Cai, Nael Abu-Ghazaleh, M. Salman Asif, Yue Dong, Amit K. Roy-Chowdhury, Chengyu Song
Abstract: Recent studies reveal that integrating new modalities into Large Language Models (LLMs), such as Vision-Language Models (VLMs), creates a new attack surface that bypasses existing safety training techniques like Supervised Fine-tuning (SFT) and Reinforcement Learning with Human Feedback (RLHF). While further SFT and RLHF-based safety training can be conducted in multi-modal settings, collecting multi-modal training datasets poses a significant challenge. Inspired by the structural design of recent multi-modal models, where, regardless of the combination of input modalities, all inputs are ultimately fused into the language space, we aim to explore whether unlearning solely in the textual domain can be effective for cross-modality safety alignment. Our evaluation across six datasets empirically demonstrates the transferability -- textual unlearning in VLMs significantly reduces the Attack Success Rate (ASR) to less than 8\% and in some cases, even as low as nearly 2\% for both text-based and vision-text-based attacks, alongside preserving the utility. Moreover, our experiments show that unlearning with a multi-modal dataset offers no potential benefits but incurs significantly increased computational demands, possibly up to 6 times higher.
Authors: Suraj Anand, David Getzen
Abstract: Numerous algorithms have been proposed to $\textit{align}$ language models to remove undesirable behaviors. However, the challenges associated with a very large state space and creating a proper reward function often result in various jailbreaks. Our paper aims to examine this effect of reward in the controlled setting of positive sentiment language generation. Instead of online training of a reward model based on human feedback, we employ a statically learned sentiment classifier. We also consider a setting where our model's weights and activations are exposed to an end-user after training. We examine a pretrained GPT-2 through the lens of mechanistic interpretability before and after proximal policy optimization (PPO) has been applied to promote positive sentiment responses. Using these insights, we (1) attempt to "hack" the PPO-ed model to generate negative sentiment responses and (2) add a term to the reward function to try and alter `negative' weights.
Authors: Louis Ledoux, Marc Casas
Abstract: We present a versatile open-source framework designed to facilitate efficient, numerically-tailored Matrix-Matrix Multiplications (MMMs). The framework offers two primary contributions: first, a fine-tuned, automated pipeline for arithmetic datapath generation, enabling highly customizable systolic MMM kernels; second, seamless integration of the generated kernels into user code, irrespective of the programming language employed, without necessitating modifications. The framework demonstrates a systematic enhancement in accuracy per energy cost across diverse High Performance Computing (HPC) workloads displaying a variety of numerical requirements, such as Artificial Intelligence (AI) inference and Sea Surface Height (SSH) computation. For AI inference, we consider a set of state-of-the-art neural network models, namely ResNet18, ResNet34, ResNet50, DenseNet121, DenseNet161, DenseNet169, and VGG11, in conjunction with two datasets, two computer formats, and 27 distinct intermediate arithmetic datapaths. Our approach consistently reduces energy consumption across all cases, with a notable example being the reduction by factors of $3.3\times$ for IEEE754-32 and $1.4\times$ for Bfloat16 during ImageNet inference with ResNet50. This is accomplished while maintaining accuracies of $82.3\%$ and $86\%$, comparable to those achieved with conventional Floating-Point Units (FPUs). In the context of SSH computation, our method achieves fully-reproducible results using double-precision words, surpassing the accuracy of conventional double- and quad-precision arithmetic in FPUs. Our approach enhances SSH computation accuracy by a minimum of $5\times$ and $27\times$ compared to IEEE754-64 and IEEE754-128, respectively, resulting in $5.6\times$ and $15.1\times$ improvements in accuracy per power cost.
Authors: Shuhong Liu, Nozomi Akashi, Qingyao Huang, Yasuo Kuniyoshi, Kohei Nakajima
Abstract: Chaos presents complex dynamics arising from nonlinearity and a sensitivity to initial states. These characteristics suggest a depth of expressivity that underscores their potential for advanced computational applications. However, strategies to effectively exploit chaotic dynamics for information processing have largely remained elusive. In this study, we reveal that the essence of chaos can be found in various state-of-the-art deep neural networks. Drawing inspiration from this revelation, we propose a novel method that directly leverages chaotic dynamics for deep learning architectures. Our approach is systematically evaluated across distinct chaotic systems. In all instances, our framework presents superior results to conventional deep neural networks in terms of accuracy, convergence speed, and efficiency. Furthermore, we found an active role of transient chaos formation in our scheme. Collectively, this study offers a new path for the integration of chaos, which has long been overlooked in information processing, and provides insights into the prospective fusion of chaotic dynamics within the domains of machine learning and neuromorphic computation.
Authors: Yihan Wu, Ruibo Chen, Zhengmian Hu, Yanshuo Chen, Junfeng Guo, Hongyang Zhang, Heng Huang
Abstract: Language model (LM) watermarking techniques inject a statistical signal into LM-generated content by substituting the random sampling process with pseudo-random sampling, using watermark keys as the random seed. Among these statistical watermarking approaches, distortion-free watermarks are particularly crucial because they embed watermarks into LM-generated content without compromising generation quality. However, one notable limitation of pseudo-random sampling compared to true-random sampling is that, under the same watermark keys (i.e., key collision), the results of pseudo-random sampling exhibit correlations. This limitation could potentially undermine the distortion-free property. Our studies reveal that key collisions are inevitable due to the limited availability of watermark keys, and existing distortion-free watermarks exhibit a significant distribution bias toward the original LM distribution in the presence of key collisions. Moreover, achieving a perfect distortion-free watermark is impossible as no statistical signal can be embedded under key collisions. To reduce the distribution bias caused by key collisions, we introduce a new family of distortion-free watermarks--beta-watermark. Experimental results support that the beta-watermark can effectively reduce the distribution bias under key collisions.
Authors: Jingjing Zheng, Xin Yuan, Kai Li, Wei Ni, Eduardo Tovar, Jon Crowcroft
Abstract: Recent attacks on federated learning (FL) can introduce malicious model updates that circumvent widely adopted Euclidean distance-based detection methods. This paper proposes a novel defense strategy, referred to as LayerCAM-AE, designed to counteract model poisoning in federated learning. The LayerCAM-AE puts forth a new Layer Class Activation Mapping (LayerCAM) integrated with an autoencoder (AE), significantly enhancing detection capabilities. Specifically, LayerCAM-AE generates a heat map for each local model update, which is then transformed into a more compact visual format. The autoencoder is designed to process the LayerCAM heat maps from the local model updates, improving their distinctiveness and thereby increasing the accuracy in spotting anomalous maps and malicious local models. To address the risk of misclassifications with LayerCAM-AE, a voting algorithm is developed, where a local model update is flagged as malicious if its heat maps are consistently suspicious over several rounds of communication. Extensive tests of LayerCAM-AE on the SVHN and CIFAR-100 datasets are performed under both Independent and Identically Distributed (IID) and non-IID settings in comparison with existing ResNet-50 and REGNETY-800MF defense models. Experimental results show that LayerCAM-AE increases detection rates (Recall: 1.0, Precision: 1.0, FPR: 0.0, Accuracy: 1.0, F1 score: 1.0, AUC: 1.0) and test accuracy in FL, surpassing the performance of both the ResNet-50 and REGNETY-800MF. Our code is available at: https://github.com/jjzgeeks/LayerCAM-AE
Authors: Joshua Collyer, Tim Watson, Iain Phillips
Abstract: Binary code similarity detection is an important problem with applications in areas like malware analysis, vulnerability research and plagiarism detection. This paper proposes a novel graph neural network architecture combined with a novel graph data representation called call graphlets. A call graphlet encodes the neighborhood around each function in a binary executable, capturing the local and global context through a series of statistical features. A specialized graph neural network model is then designed to operate on this graph representation, learning to map it to a feature vector that encodes semantic code similarities using deep metric learning. The proposed approach is evaluated across four distinct datasets covering different architectures, compiler toolchains, and optimization levels. Experimental results demonstrate that the combination of call graphlets and the novel graph neural network architecture achieves state-of-the-art performance compared to baseline techniques across cross-architecture, mono-architecture and zero shot tasks. In addition, our proposed approach also performs well when evaluated against an out-of-domain function inlining task. Overall, the work provides a general and effective graph neural network-based solution for conducting binary code similarity detection.
Authors: Varun Reddy
Abstract: PPINtonus is a system for the early detection of Parkinson's Disease (PD) utilizing deep-learning tonal analysis, providing a cost-effective and accessible alternative to traditional neurological examinations. Partnering with the Parkinson's Voice Project (PVP), PPINtonus employs a semi-supervised conditional generative adversarial network to generate synthetic data points, enhancing the training dataset for a multi-layered deep neural network. Combined with PRAAT phonetics software, this network accurately assesses biomedical voice measurement values from a simple 120-second vocal test performed with a standard microphone in typical household noise conditions. The model's performance was validated using a confusion matrix, achieving an impressive 92.5 \% accuracy with a low false negative rate. PPINtonus demonstrated a precision of 92.7 \%, making it a reliable tool for early PD detection. The non-intrusive and efficient methodology of PPINtonus can significantly benefit developing countries by enabling early diagnosis and improving the quality of life for millions of PD patients through timely intervention and management.
Authors: Li Wang, Xiangzheng Fu, Jiahao Yang, Xinyi Zhang, Xiucai Ye, Yiping Liu, Tetsuya Sakurai, Xiangxiang Zeng
Abstract: Deep learning holds a big promise for optimizing existing peptides with more desirable properties, a critical step towards accelerating new drug discovery. Despite the recent emergence of several optimized Antimicrobial peptides(AMP) generation methods, multi-objective optimizations remain still quite challenging for the idealism-realism tradeoff. Here, we establish a multi-objective AMP synthesis pipeline (MoFormer) for the simultaneous optimization of multi-attributes of AMPs. MoFormer improves the desired attributes of AMP sequences in a highly structured latent space, guided by conditional constraints and fine-grained multi-descriptor.We show that MoFormer outperforms existing methods in the generation task of enhanced antimicrobial activity and minimal hemolysis. We also utilize a Pareto-based non-dominated sorting algorithm and proxies based on large model fine-tuning to hierarchically rank the candidates. We demonstrate substantial property improvement using MoFormer from two perspectives: (1) employing molecular simulations and scoring interactions among amino acids to decipher the structure and functionality of AMPs; (2) visualizing latent space to examine the qualities and distribution features, verifying an effective means to facilitate multi-objective optimization AMPs with design constraints
Authors: Andis Draguns, Andrew Gritsevskiy, Sumeet Ramesh Motwani, Charlie Rogers-Smith, Jeffrey Ladish, Christian Schroeder de Witt
Abstract: The rapid proliferation of open-source language models significantly increases the risks of downstream backdoor attacks. These backdoors can introduce dangerous behaviours during model deployment and can evade detection by conventional cybersecurity monitoring systems. In this paper, we introduce a novel class of backdoors in autoregressive transformer models, that, in contrast to prior art, are unelicitable in nature. Unelicitability prevents the defender from triggering the backdoor, making it impossible to evaluate or detect ahead of deployment even if given full white-box access and using automated techniques, such as red-teaming or certain formal verification methods. We show that our novel construction is not only unelicitable thanks to using cryptographic techniques, but also has favourable robustness properties. We confirm these properties in empirical investigations, and provide evidence that our backdoors can withstand state-of-the-art mitigation strategies. Additionally, we expand on previous work by showing that our universal backdoors, while not completely undetectable in white-box settings, can be harder to detect than some existing designs. By demonstrating the feasibility of seamlessly integrating backdoors into transformer models, this paper fundamentally questions the efficacy of pre-deployment detection strategies. This offers new insights into the offence-defence balance in AI safety and security.
Authors: Max Hopkins, Russell Impagliazzo, Daniel Kane, Sihan Liu, Christopher Ye
Abstract: The replicability crisis is a major issue across nearly all areas of empirical science, calling for the formal study of replicability in statistics. Motivated in this context, [Impagliazzo, Lei, Pitassi, and Sorrell STOC 2022] introduced the notion of replicable learning algorithms, and gave basic procedures for $1$-dimensional tasks including statistical queries. In this work, we study the computational and statistical cost of replicability for several fundamental high dimensional statistical tasks, including multi-hypothesis testing and mean estimation. Our main contribution establishes a computational and statistical equivalence between optimal replicable algorithms and high dimensional isoperimetric tilings. As a consequence, we obtain matching sample complexity upper and lower bounds for replicable mean estimation of distributions with bounded covariance, resolving an open problem of [Bun, Gaboardi, Hopkins, Impagliazzo, Lei, Pitassi, Sivakumar, and Sorrell, STOC2023] and for the $N$-Coin Problem, resolving a problem of [Karbasi, Velegkas, Yang, and Zhou, NeurIPS2023] up to log factors. While our equivalence is computational, allowing us to shave log factors in sample complexity from the best known efficient algorithms, efficient isoperimetric tilings are not known. To circumvent this, we introduce several relaxed paradigms that do allow for sample and computationally efficient algorithms, including allowing pre-processing, adaptivity, and approximate replicability. In these cases we give efficient algorithms matching or beating the best known sample complexity for mean estimation and the coin problem, including a generic procedure that reduces the standard quadratic overhead of replicability to linear in expectation.
Authors: Shijin Duan, Chenghong Wang, Hongwu Peng, Yukui Luo, Wujie Wen, Caiwen Ding, Xiaolin Xu
Abstract: As privacy-preserving becomes a pivotal aspect of deep learning (DL) development, multi-party computation (MPC) has gained prominence for its efficiency and strong security. However, the practice of current MPC frameworks is limited, especially when dealing with large neural networks, exemplified by the prolonged execution time of 25.8 seconds for secure inference on ResNet-152. The primary challenge lies in the reliance of current MPC approaches on additive secret sharing, which incurs significant communication overhead with non-linear operations such as comparisons. Furthermore, additive sharing suffers from poor scalability on party size. In contrast, the evolving landscape of MPC necessitates accommodating a larger number of compute parties and ensuring robust performance against malicious activities or computational failures. In light of these challenges, we propose SSNet, which for the first time, employs Shamir's secret sharing (SSS) as the backbone of MPC-based ML framework. We meticulously develop all framework primitives and operations for secure DL models tailored to seamlessly integrate with the SSS scheme. SSNet demonstrates the ability to scale up party numbers straightforwardly and embeds strategies to authenticate the computation correctness without incurring significant performance overhead. Additionally, SSNet introduces masking strategies designed to reduce communication overhead associated with non-linear operations. We conduct comprehensive experimental evaluations on commercial cloud computing infrastructure from Amazon AWS, as well as across diverse prevalent DNN models and datasets. SSNet demonstrates a substantial performance boost, achieving speed-ups ranging from 3x to 14x compared to SOTA MPC frameworks. Moreover, SSNet also represents the first framework that is evaluated on a five-party computation setup, in the context of secure DL inference.
Authors: Fernando Martinez, Mariyam Mapkar, Ali Alfatemi, Mohamed Rahouti, Yufeng Xin, Kaiqi Xiong, Nasir Ghani
Abstract: Distributed Denial of Service (DDoS) attacks pose an increasingly substantial cybersecurity threat to organizations across the globe. In this paper, we introduce a new deep learning-based technique for detecting DDoS attacks, a paramount cybersecurity challenge with evolving complexity and scale. Specifically, we propose a new dual-space prototypical network that leverages a unique dual-space loss function to enhance detection accuracy for various attack patterns through geometric and angular similarity measures. This approach capitalizes on the strengths of representation learning within the latent space (a lower-dimensional representation of data that captures complex patterns for machine learning analysis), improving the model's adaptability and sensitivity towards varying DDoS attack vectors. Our comprehensive evaluation spans multiple training environments, including offline training, simulated online training, and prototypical network scenarios, to validate the model's robustness under diverse data abundance and scarcity conditions. The Multilayer Perceptron (MLP) with Attention, trained with our dual-space prototypical design over a reduced training set, achieves an average accuracy of 94.85% and an F1-Score of 94.71% across our tests, showcasing its effectiveness in dynamic and constrained real-world scenarios.
Authors: Noah Golowich, Ankur Moitra
Abstract: Motivated by the problem of detecting AI-generated text, we consider the problem of watermarking the output of language models with provable guarantees. We aim for watermarks which satisfy: (a) undetectability, a cryptographic notion introduced by Christ, Gunn & Zamir (2024) which stipulates that it is computationally hard to distinguish watermarked language model outputs from the model's actual output distribution; and (b) robustness to channels which introduce a constant fraction of adversarial insertions, substitutions, and deletions to the watermarked text. Earlier schemes could only handle stochastic substitutions and deletions, and thus we are aiming for a more natural and appealing robustness guarantee that holds with respect to edit distance. Our main result is a watermarking scheme which achieves both undetectability and robustness to edits when the alphabet size for the language model is allowed to grow as a polynomial in the security parameter. To derive such a scheme, we follow an approach introduced by Christ & Gunn (2024), which proceeds via first constructing pseudorandom codes satisfying undetectability and robustness properties analogous to those above; our key idea is to handle adversarial insertions and deletions by interpreting the symbols as indices into the codeword, which we call indexing pseudorandom codes. Additionally, our codes rely on weaker computational assumptions than used in previous work. Then we show that there is a generic transformation from such codes over large alphabets to watermarking schemes for arbitrary language models.
Authors: Vladimir Fanaskov, Tianchi Yu, Alexander Rudikov, Ivan Oseledets
Abstract: The primal approach to physics-informed learning is a residual minimization. We argue that residual is, at best, an indirect measure of the error of approximate solution and propose to train with error majorant instead. Since error majorant provides a direct upper bound on error, one can reliably estimate how close PiNN is to the exact solution and stop the optimization process when the desired accuracy is reached. We call loss function associated with error majorant $\textbf{Astral}$: neur$\textbf{A}$l a po$\textbf{ST}$erio$\textbf{RI}$ function$\textbf{A}$l Loss. To compare Astral and residual loss functions, we illustrate how error majorants can be derived for various PDEs and conduct experiments with diffusion equations (including anisotropic and in the L-shaped domain), convection-diffusion equation, temporal discretization of Maxwell's equation, and magnetostatics problem. The results indicate that Astral loss is competitive to the residual loss, typically leading to faster convergence and lower error (e.g., for Maxwell's equations, we observe an order of magnitude better relative error and training time). We also report that the error estimate obtained with Astral loss is usually tight enough to be informative, e.g., for a highly anisotropic equation, on average, Astral overestimates error by a factor of $1.5$, and for convection-diffusion by a factor of $1.7$.
Authors: Aviv Shamsian, Aviv Navon, Neta Glazer, Gill Hetz, Joseph Keshet
Abstract: Automatic Speech Recognition (ASR) technology has made significant progress in recent years, providing accurate transcription across various domains. However, some challenges remain, especially in noisy environments and specialized jargon. In this paper, we propose a novel approach for improved jargon word recognition by contextual biasing Whisper-based models. We employ a keyword spotting model that leverages the Whisper encoder representation to dynamically generate prompts for guiding the decoder during the transcription process. We introduce two approaches to effectively steer the decoder towards these prompts: KG-Whisper, which is aimed at fine-tuning the Whisper decoder, and KG-Whisper-PT, which learns a prompt prefix. Our results show a significant improvement in the recognition accuracy of specified keywords and in reducing the overall word error rates. Specifically, in unseen language generalization, we demonstrate an average WER improvement of 5.1% over Whisper.
Authors: Arnav Kundu, Prateeth Nayak, Hywel Richards, Priyanka Padmanabhan, Devang Naik
Abstract: Always-on machine learning models require a very low memory and compute footprint. Their restricted parameter count limits the model's capacity to learn, and the effectiveness of the usual training algorithms to find the best parameters. Here we show that a small convolutional model can be better trained by first refactoring its computation into a larger redundant multi-branched architecture. Then, for inference, we algebraically re-parameterize the trained model into the single-branched form with fewer parameters for a lower memory footprint and compute cost. Using this technique, we show that our always-on wake-word detector model, RepCNN, provides a good trade-off between latency and accuracy during inference. RepCNN re-parameterized models are 43% more accurate than a uni-branch convolutional model while having the same runtime. RepCNN also meets the accuracy of complex architectures like BC-ResNet, while having 2x lesser peak memory usage and 10x faster runtime.
Authors: Reza Babaei, Samuel Cheng, Theresa Thai, Shangqing Zhao
Abstract: Despite the advances in medicine, cancer has remained a formidable challenge. Particularly in the case of pancreatic tumors, characterized by their diversity and late diagnosis, early detection poses a significant challenge crucial for effective treatment. The advancement of deep learning techniques, particularly supervised algorithms, has significantly propelled pancreatic tumor detection in the medical field. However, supervised deep learning approaches necessitate extensive labeled medical images for training, yet acquiring such annotations is both limited and costly. Conversely, weakly supervised anomaly detection methods, requiring only image-level annotations, have garnered interest. Existing methodologies predominantly hinge on generative adversarial networks (GANs) or autoencoder models, which can pose complexity in training and, these models may face difficulties in accurately preserving fine image details. This research presents a novel approach to pancreatic tumor detection, employing weak supervision anomaly detection through denoising diffusion algorithms. By incorporating a deterministic iterative process of adding and removing noise along with classifier guidance, the method enables seamless translation of images between diseased and healthy subjects, resulting in detailed anomaly maps without requiring complex training protocols and segmentation masks. This study explores denoising diffusion models as a recent advancement over traditional generative models like GANs, contributing to the field of pancreatic tumor detection. Recognizing the low survival rates of pancreatic cancer, this study emphasizes the need for continued research to leverage diffusion models' efficiency in medical segmentation tasks.
Authors: John Musgrave, Anca Ralescu
Abstract: Feature resolution impacts the ability of classifiers to make explainable inferences when applied to malware classification. We explore classification based on features constructed from data dependency graphs, and present results from k-Nearest Neighbors (kNN) classifiers. Our study demonstrates that classification based on a novel feature representation not only yields high accuracy, but also increases explainability in inference, as features of data dependency are directly representative of program behavior. We present classification results using the Microsoft Kaggle 2015 malware dataset which was processed with a novel approach to feature extraction and representation. We show that non-parametric approaches to classification in the metric space are able to obtain classification accuracy of 87.5\% when applied to multi-class classification in the Kaggle malware dataset. Additionally, similarity in the metric space can be calculated directly without prior training. Our results provide evidence that data dependency graphs accurately capture both semantic and structural information.
Authors: Namgyu Ho, Sangmin Bae, Taehyeon Kim, Hyunjik Jo, Yireun Kim, Tal Schuster, Adam Fisch, James Thorne, Se-Young Yun
Abstract: This paper presents the Block Transformer architecture which adopts hierarchical global-to-local modeling to autoregressive transformers to mitigate the inference bottlenecks of self-attention. To apply self-attention, the key-value (KV) cache of all previous sequences must be retrieved from memory at every decoding step. Thereby, this KV cache IO becomes a significant bottleneck in batch inference. We notice that these costs stem from applying self-attention on the global context, therefore we isolate the expensive bottlenecks of global modeling to lower layers and apply fast local modeling in upper layers. To mitigate the remaining costs in the lower layers, we aggregate input tokens into fixed size blocks and then apply self-attention at this coarse level. Context information is aggregated into a single embedding to enable upper layers to decode the next block of tokens, without global attention. Free of global attention bottlenecks, the upper layers can fully utilize the compute hardware to maximize inference throughput. By leveraging global and local modules, the Block Transformer architecture demonstrates 10-20x gains in inference throughput compared to vanilla transformers with equivalent perplexity. Our work introduces a new approach to optimize language model inference through novel application of global-to-local modeling. Code is available at https://github.com/itsnamgyu/block-transformer.
Authors: Itay Lavie, Zohar Ringel
Abstract: Kernel ridge regression (KRR) and Gaussian processes (GPs) are fundamental tools in statistics and machine learning with recent applications to highly over-parameterized deep neural networks. The ability of these tools to learn a target function is directly related to the eigenvalues of their kernel sampled on the input data. Targets having support on higher eigenvalues are more learnable. While kernels are often highly symmetric objects, the data is often not. Thus kernel symmetry seems to have little to no bearing on the above eigenvalues or learnability, making spectral analysis on real-world data challenging. Here, we show that contrary to this common lure, one may use eigenvalues and eigenfunctions associated with highly idealized data-measures to bound learnability on realistic data. As a demonstration, we give a theoretical lower bound on the sample complexity of copying heads for kernels associated with generic transformers acting on natural language.
Authors: Zoe De Simone, Sayandeep Biswas, Oscar Wu
Abstract: Window to Wall Ratios (WWR) are key to assessing the energy, daylight and ventilation performance of buildings. Studies have shown that window area has a large impact on building performance and simulation. However, data to set up these environmental models and simulations is typically not available. Instead, a standard 40% WWR is typically assumed for all buildings. This paper leverages existing computer vision window detection methods to predict WWR of buildings from external street view images using semantic segmentation, demonstrating the potential for adapting established computer vision technique in architectural applications
Authors: Surbhi Goel, Abhishek Shetty, Konstantinos Stavropoulos, Arsen Vasilyan
Abstract: We study the problem of learning under arbitrary distribution shift, where the learner is trained on a labeled set from one distribution but evaluated on a different, potentially adversarially generated test distribution. We focus on two frameworks: PQ learning [Goldwasser, A. Kalai, Y. Kalai, Montasser NeurIPS 2020], allowing abstention on adversarially generated parts of the test distribution, and TDS learning [Klivans, Stavropoulos, Vasilyan COLT 2024], permitting abstention on the entire test distribution if distribution shift is detected. All prior known algorithms either rely on learning primitives that are computationally hard even for simple function classes, or end up abstaining entirely even in the presence of a tiny amount of distribution shift. We address both these challenges for natural function classes, including intersections of halfspaces and decision trees, and standard training distributions, including Gaussians. For PQ learning, we give efficient learning algorithms, while for TDS learning, our algorithms can tolerate moderate amounts of distribution shift. At the core of our approach is an improved analysis of spectral outlier-removal techniques from learning with nasty noise. Our analysis can (1) handle arbitrarily large fraction of outliers, which is crucial for handling arbitrary distribution shifts, and (2) obtain stronger bounds on polynomial moments of the distribution after outlier removal, yielding new insights into polynomial regression under distribution shifts. Lastly, our techniques lead to novel results for tolerant testable learning [Rubinfeld and Vasilyan STOC 2023], and learning with nasty noise.
Authors: Yixuan Liu, Li Xiong, Yuhan Liu, Yujie Gu, Ruixuan Liu, Hong Chen
Abstract: Differentially Private Stochastic Gradients Descent (DP-SGD) is a prominent paradigm for preserving privacy in deep learning. It ensures privacy by perturbing gradients with random noise calibrated to their entire norm at each training step. However, this perturbation suffers from a sub-optimal performance: it repeatedly wastes privacy budget on the general converging direction shared among gradients from different batches, which we refer as common knowledge, yet yields little information gain. Motivated by this, we propose a differentially private training framework with early gradient decomposition and reconstruction (DPDR), which enables more efficient use of the privacy budget. In essence, it boosts model utility by focusing on incremental information protection and recycling the privatized common knowledge learned from previous gradients at early training steps. Concretely, DPDR incorporates three steps. First, it disentangles common knowledge and incremental information in current gradients by decomposing them based on previous noisy gradients. Second, most privacy budget is spent on protecting incremental information for higher information gain. Third, the model is updated with the gradient reconstructed from recycled common knowledge and noisy incremental information. Theoretical analysis and extensive experiments show that DPDR outperforms state-of-the-art baselines on both convergence rate and accuracy.
Authors: Dehong Xu, Liang Qiu, Minseok Kim, Faisal Ladhak, Jaeyoung Do
Abstract: Pre-trained large-scale language models (LLMs) excel at producing coherent articles, yet their outputs may be untruthful, toxic, or fail to align with user expectations. Current approaches focus on using reinforcement learning with human feedback (RLHF) to improve model alignment, which works by transforming coarse human preferences of LLM outputs into a feedback signal that guides the model learning process. However, because this approach operates on sequence-level feedback, it lacks the precision to identify the exact parts of the output affecting user preferences. To address this gap, we propose a method to enhance LLM alignment through fine-grained token-level supervision. Specifically, we ask annotators to minimally edit less preferred responses within the standard reward modeling dataset to make them more favorable, ensuring changes are made only where necessary while retaining most of the original content. The refined dataset is used to train a token-level reward model, which is then used for training our fine-grained Proximal Policy Optimization (PPO) model. Our experiment results demonstrate that this approach can achieve up to an absolute improvement of $5.1\%$ in LLM performance, in terms of win rate against the reference model, compared with the traditional PPO model.
Authors: Wayner Barrios, SouYoung Jin
Abstract: While the Self-Attention mechanism in the Transformer model has proven to be effective in many domains, we observe that it is less effective in more diverse settings (e.g. multimodality) due to the varying granularity of each token and the high computational demands of lengthy sequences. To address the challenges, we introduce the Learnable Attention Mask (LAM), strategically designed to globally regulate attention maps and prioritize critical tokens within the sequence. Leveraging the Self-Attention module in a BERT-like transformer network, our approach adeptly captures associations between tokens. The extension of the LAM to a multi-layer version accommodates the varied information aspects embedded at each layer of the Transformer network. Comprehensive experimental validation on various datasets, such as MADv2, QVHighlights, ImageNet 1K, and MSRVTT, demonstrates the efficacy of the LAM, exemplifying its ability to enhance model performance while mitigating redundant computations. This pioneering approach presents a significant advancement in enhancing the understanding of complex scenarios, such as in movie understanding.
Authors: Sigur de Vries, Sander Keemink, Marcel van Gerven
Abstract: Artificial intelligence (AI) techniques are increasingly being applied to solve control problems. However, control systems developed in AI are often black-box methods, in that it is not clear how and why they generate their outputs. A lack of transparency can be problematic for control tasks in particular, because it complicates the identification of biases or errors, which in turn negatively influences the user's confidence in the system. To improve the interpretability and transparency in control systems, the black-box structure can be replaced with white-box symbolic policies described by mathematical expressions. Genetic programming offers a gradient-free method to optimise the structure of non-differentiable mathematical expressions. In this paper, we show that genetic programming can be used to discover symbolic control systems. This is achieved by learning a symbolic representation of a function that transforms observations into control signals. We consider both systems that implement static control policies without memory and systems that implement dynamic memory-based control policies. In case of the latter, the discovered function becomes the state equation of a differential equation, which allows for evidence integration. Our results show that symbolic policies are discovered that perform comparably with black-box policies on a variety of control tasks. Furthermore, the additional value of the memory capacity in the dynamic policies is demonstrated on experiments where static policies fall short. Overall, we demonstrate that white-box symbolic policies can be optimised with genetic programming, while offering interpretability and transparency that lacks in black-box models.
Authors: Chiraag Kaushik, Justin Romberg, Vidya Muthukumar
Abstract: The classical iteratively reweighted least-squares (IRLS) algorithm aims to recover an unknown signal from linear measurements by performing a sequence of weighted least squares problems, where the weights are recursively updated at each step. Varieties of this algorithm have been shown to achieve favorable empirical performance and theoretical guarantees for sparse recovery and $\ell_p$-norm minimization. Recently, some preliminary connections have also been made between IRLS and certain types of non-convex linear neural network architectures that are observed to exploit low-dimensional structure in high-dimensional linear models. In this work, we provide a unified asymptotic analysis for a family of algorithms that encompasses IRLS, the recently proposed lin-RFM algorithm (which was motivated by feature learning in neural networks), and the alternating minimization algorithm on linear diagonal neural networks. Our analysis operates in a "batched" setting with i.i.d. Gaussian covariates and shows that, with appropriately chosen reweighting policy, the algorithm can achieve favorable performance in only a handful of iterations. We also extend our results to the case of group-sparse recovery and show that leveraging this structure in the reweighting scheme provably improves test error compared to coordinate-wise reweighting.
Authors: Samuel Scheele, Katherine Picchione, Jeffrey Liu
Abstract: ML-based computer vision models are promising tools for supporting emergency management operations following natural disasters. Arial photographs taken from small manned and unmanned aircraft can be available soon after a disaster and provide valuable information from multiple perspectives for situational awareness and damage assessment applications. However, emergency managers often face challenges finding the most relevant photos among the tens of thousands that may be taken after an incident. While ML-based solutions could enable more effective use of aerial photographs, there is still a lack of training data for imagery of this type from multiple perspectives and for multiple hazard types. To address this, we present the LADI v2 (Low Altitude Disaster Imagery version 2) dataset, a curated set of about 10,000 disaster images captured in the United States by the Civil Air Patrol (CAP) in response to federally-declared emergencies (2015-2023) and annotated for multi-label classification by trained CAP volunteers. We also provide two pretrained baseline classifiers and compare their performance to state-of-the-art vision-language models in multi-label classification. The data and code are released publicly to support the development of computer vision models for emergency management research and applications.
Authors: Berthy T. Feng, Katherine L. Bouman, William T. Freeman
Abstract: Reconstructing images from the Event Horizon Telescope (EHT) observations of M87*, the supermassive black hole at the center of the galaxy M87, depends on a prior to impose desired image statistics. However, given the impossibility of directly observing black holes, there is no clear choice for a prior. We present a framework for flexibly designing a range of priors, each bringing different biases to the image reconstruction. These priors can be weak (e.g., impose only basic natural-image statistics) or strong (e.g., impose assumptions of black-hole structure). Our framework uses Bayesian inference with score-based priors, which are data-driven priors arising from a deep generative model that can learn complicated image distributions. Using our Bayesian imaging approach with sophisticated data-driven priors, we can assess how visual features and uncertainty of reconstructed images change depending on the prior. In addition to simulated data, we image the real EHT M87* data and discuss how recovered features are influenced by the choice of prior.
Authors: Wenyue Hua, Kaijie Zhu, Lingyao Li, Lizhou Fan, Shuhang Lin, Mingyu Jin, Haochen Xue, Zelong Li, JinDong Wang, Yongfeng Zhang
Abstract: This study intends to systematically disentangle pure logic reasoning and text understanding by investigating the contrast across abstract and contextualized logical problems from a comprehensive set of domains. We explore whether LLMs demonstrate genuine reasoning capabilities across various domains when the underlying logical structure remains constant. We focus on two main questions (1) Can abstract logical problems alone accurately benchmark an LLM's reasoning ability in real-world scenarios, disentangled from contextual support in practical settings? (2) Does fine-tuning LLMs on abstract logic problem generalize to contextualized logic problems and vice versa? To investigate these questions, we focus on standard propositional logic, specifically propositional deductive and abductive logic reasoning. In particular, we construct instantiated datasets for deductive and abductive reasoning with 4 levels of difficulty, encompassing 12 distinct categories or domains based on the categorization of Wikipedia. Our experiments aim to provide insights into disentangling context in logical reasoning and the true reasoning capabilities of LLMs and their generalization potential. The code and dataset are available at: https://github.com/agiresearch/ContextHub.
Authors: Hilal Asi, Daogao Liu, Kevin Tian
Abstract: We study the problem of differentially private stochastic convex optimization (DP-SCO) with heavy-tailed gradients, where we assume a $k^{\text{th}}$-moment bound on the Lipschitz constants of sample functions rather than a uniform bound. We propose a new reduction-based approach that enables us to obtain the first optimal rates (up to logarithmic factors) in the heavy-tailed setting, achieving error $G_2 \cdot \frac 1 {\sqrt n} + G_k \cdot (\frac{\sqrt d}{n\epsilon})^{1 - \frac 1 k}$ under $(\epsilon, \delta)$-approximate differential privacy, up to a mild $\textup{polylog}(\frac{1}{\delta})$ factor, where $G_2^2$ and $G_k^k$ are the $2^{\text{nd}}$ and $k^{\text{th}}$ moment bounds on sample Lipschitz constants, nearly-matching a lower bound of [Lowy and Razaviyayn 2023]. We further give a suite of private algorithms in the heavy-tailed setting which improve upon our basic result under additional assumptions, including an optimal algorithm under a known-Lipschitz constant assumption, a near-linear time algorithm for smooth functions, and an optimal linear time algorithm for smooth generalized linear models.
Authors: Fran\c{c}ois Roewer-Despr\'es, Jinyue Feng, Zining Zhu, Frank Rudzicz
Abstract: We present $\texttt{ACCORD}$, a framework and benchmark suite for disentangling the commonsense grounding and reasoning abilities of large language models (LLMs) through controlled, multi-hop counterfactuals. $\texttt{ACCORD}$ introduces formal elements to commonsense reasoning to explicitly control and quantify reasoning complexity beyond the typical 1 or 2 hops. Uniquely, $\texttt{ACCORD}$ can automatically generate benchmarks of arbitrary reasoning complexity, and so it scales with future LLM improvements. Benchmarking state-of-the-art LLMs -- including GPT-4o (2024-05-13), Llama-3-70B-Instruct, and Mixtral-8x22B-Instruct-v0.1 -- shows performance degrading to random chance with only moderate scaling, leaving substantial headroom for improvement. We release a leaderboard of the benchmark suite tested in this work, as well as code for automatically generating more complex benchmarks.
Authors: Kiymet Akdemir, Pinar Yanardag
Abstract: Text-to-image diffusion models have recently taken center stage as pivotal tools in promoting visual creativity across an array of domains such as comic book artistry, children's literature, game development, and web design. These models harness the power of artificial intelligence to convert textual descriptions into vivid images, thereby enabling artists and creators to bring their imaginative concepts to life with unprecedented ease. However, one of the significant hurdles that persist is the challenge of maintaining consistency in character generation across diverse contexts. Variations in textual prompts, even if minor, can yield vastly different visual outputs, posing a considerable problem in projects that require a uniform representation of characters throughout. In this paper, we introduce a novel framework designed to produce consistent character representations from a single text prompt across diverse settings. Through both quantitative and qualitative analyses, we demonstrate that our framework outperforms existing methods in generating characters with consistent visual identities, underscoring its potential to transform creative industries. By addressing the critical challenge of character consistency, we not only enhance the practical utility of these models but also broaden the horizons for artistic and creative expression.
Authors: Yu-Wen Chen, Julia Hirschberg
Abstract: Summarizing medical conversations poses unique challenges due to the specialized domain and the difficulty of collecting in-domain training data. In this study, we investigate the performance of state-of-the-art doctor-patient conversation generative summarization models on the out-of-domain data. We divide the summarization model of doctor-patient conversation into two configurations: (1) a general model, without specifying subjective (S), objective (O), and assessment (A) and plan (P) notes; (2) a SOAP-oriented model that generates a summary with SOAP sections. We analyzed the limitations and strengths of the fine-tuning language model-based methods and GPTs on both configurations. We also conducted a Linguistic Inquiry and Word Count analysis to compare the SOAP notes from different datasets. The results exhibit a strong correlation for reference notes across different datasets, indicating that format mismatch (i.e., discrepancies in word distribution) is not the main cause of performance decline on out-of-domain data. Lastly, a detailed analysis of SOAP notes is included to provide insights into missing information and hallucinations introduced by the models.
Authors: Firas Trabelsi, David Vilar, Mara Finkelstein, Markus Freitag
Abstract: Minimum Bayes Risk (MBR) decoding is a powerful decoding strategy widely used for text generation tasks, but its quadratic computational complexity limits its practical application. This paper presents a novel approach for approximating MBR decoding using matrix completion techniques, focusing on the task of machine translation. We formulate MBR decoding as a matrix completion problem, where the utility metric scores between candidate hypotheses and pseudo-reference translations form a low-rank matrix. First, we empirically show that the scores matrices indeed have a low-rank structure. Then, we exploit this by only computing a random subset of the scores and efficiently recover the missing entries in the matrix by applying the Alternating Least Squares (ALS) algorithm, thereby enabling a fast approximation of the MBR decoding process. Our experimental results on machine translation tasks demonstrate that the proposed method requires 1/16 utility metric computations compared to vanilla MBR decoding while achieving equal translation quality measured by COMET22 on the WMT22 dataset (en<>de and en<>ru). We also benchmark our method against other approximation methods and we show gains in quality when comparing to them.
Authors: Dominic B. Dayta
Abstract: We introduce YOASOVI, an algorithm for performing fast, self-correcting stochastic optimization for Variational Inference (VI) on large Bayesian heirarchical models. To accomplish this, we take advantage of available information on the objective function used for stochastic VI at each iteration and replace regular Monte Carlo sampling with acceptance sampling. Rather than spend computational resources drawing and evaluating over a large sample for the gradient, we draw only one sample and accept it with probability proportional to the expected improvement in the objective. The following paper develops two versions of the algorithm: the first one based on a naive intuition, and another building up the algorithm as a Metropolis-type scheme. Empirical results based on simulations and benchmark datasets for multivariate Gaussian mixture models show that YOASOVI consistently converges faster (in clock time) and within better optimal neighborhoods than both regularized Monte Carlo and Quasi-Monte Carlo VI algorithms.
Authors: Ryo Yonetani
Abstract: This paper presents TSPDiffuser, a novel data-driven path planner for traveling salesperson path planning problems (TSPPPs) in environments rich with obstacles. Given a set of destinations within obstacle maps, our objective is to efficiently find the shortest possible collision-free path that visits all the destinations. In TSPDiffuser, we train a diffusion model on a large collection of TSPPP instances and their respective solutions to generate plausible paths for unseen problem instances. The model can then be employed as a learned sampler to construct a roadmap that contains potential solutions with a small number of nodes and edges. This approach enables efficient and accurate estimation of traveling costs between destinations, effectively addressing the primary computational challenge in solving TSPPPs. Experimental evaluations with diverse synthetic and real-world indoor/outdoor environments demonstrate the effectiveness of TSPDiffuser over existing methods in terms of the trade-off between solution quality and computational time requirements.
Authors: Ilker Demirel, Ahmed Alaa, Anthony Philippakis, David Sontag
Abstract: Causal inferences from a randomized controlled trial (RCT) may not pertain to a target population where some effect modifiers have a different distribution. Prior work studies generalizing the results of a trial to a target population with no outcome but covariate data available. We show how the limited size of trials makes generalization a statistically infeasible task, as it requires estimating complex nuisance functions. We develop generalization algorithms that supplement the trial data with a prediction model learned from an additional observational study (OS), without making any assumptions on the OS. We theoretically and empirically show that our methods facilitate better generalization when the OS is high-quality, and remain robust when it is not, and e.g., have unmeasured confounding.
Authors: Yuchen Zhuang, Haotian Sun, Yue Yu, Qifan Wang, Chao Zhang, Bo Dai
Abstract: Personalization has emerged as a critical research area in modern intelligent systems, focusing on mining users' behavioral history and adapting to their preferences for delivering tailored experiences. Despite the remarkable few-shot capabilities exhibited by black-box large language models (LLMs), the inherent opacity of their model parameters presents significant challenges in aligning the generated output with individual expectations. Existing solutions have primarily focused on prompt design to incorporate user-specific profiles and behaviors; however, such approaches often struggle to generalize effectively due to their inability to capture shared knowledge among all users. To address these challenges, we propose HYDRA, a model factorization framework that captures both user-specific behavior patterns from historical data and shared general knowledge among all users to deliver personalized generation. In order to capture user-specific behavior patterns, we first train a reranker to prioritize the most useful information from top-retrieved relevant historical records. By combining the prioritized history with the corresponding query, we train an adapter to align the output with individual user-specific preferences, eliminating the reliance on access to inherent model parameters of black-box LLMs. Both the reranker and the adapter can be decomposed into a base model with multiple user-specific heads, resembling a hydra. The base model maintains shared knowledge across users, while the multiple personal heads capture user-specific preferences. Experimental results demonstrate that HYDRA outperforms existing state-of-the-art prompt-based methods by an average relative improvement of 9.01% across five diverse personalization tasks in the LaMP benchmark. Our implementation is available at https://github.com/night-chen/HYDRA.
Authors: Dom Huh, Prasant Mohapatra
Abstract: Sample efficiency remains a key challenge in multi-agent reinforcement learning (MARL). A promising approach is to learn a meaningful latent representation space through auxiliary learning objectives alongside the MARL objective to aid in learning a successful control policy. In our work, we present MAPO-LSO (Multi-Agent Policy Optimization with Latent Space Optimization) which applies a form of comprehensive representation learning devised to supplement MARL training. Specifically, MAPO-LSO proposes a multi-agent extension of transition dynamics reconstruction and self-predictive learning that constructs a latent state optimization scheme that can be trivially extended to current state-of-the-art MARL algorithms. Empirical results demonstrate MAPO-LSO to show notable improvements in sample efficiency and learning performance compared to its vanilla MARL counterpart without any additional MARL hyperparameter tuning on a diverse suite of MARL tasks.
Authors: Haike Xu, Sandeep Silwal, Piotr Indyk
Abstract: We propose a new "bi-metric" framework for designing nearest neighbor data structures. Our framework assumes two dissimilarity functions: a ground-truth metric that is accurate but expensive to compute, and a proxy metric that is cheaper but less accurate. In both theory and practice, we show how to construct data structures using only the proxy metric such that the query procedure achieves the accuracy of the expensive metric, while only using a limited number of calls to both metrics. Our theoretical results instantiate this framework for two popular nearest neighbor search algorithms: DiskANN and Cover Tree. In both cases we show that, as long as the proxy metric used to construct the data structure approximates the ground-truth metric up to a bounded factor, our data structure achieves arbitrarily good approximation guarantees with respect to the ground-truth metric. On the empirical side, we apply the framework to the text retrieval problem with two dissimilarity functions evaluated by ML models with vastly different computational costs. We observe that for almost all data sets in the MTEB benchmark, our approach achieves a considerably better accuracy-efficiency tradeoff than the alternatives, such as re-ranking.
Authors: Jinhao Li, Haopeng Li, Sarah Erfani, Lei Feng, James Bailey, Feng Liu
Abstract: It has recently been discovered that using a pre-trained vision-language model (VLM), e.g., CLIP, to align a whole query image with several finer text descriptions generated by a large language model can significantly enhance zero-shot performance. However, in this paper, we empirically find that the finer descriptions tend to align more effectively with local areas of the query image rather than the whole image, and then we theoretically validate this finding. Thus, we present a method called weighted visual-text cross alignment (WCA). This method begins with a localized visual prompting technique, designed to identify local visual areas within the query image. The local visual areas are then cross-aligned with the finer descriptions by creating a similarity matrix using the pre-trained VLM. To determine how well a query image aligns with each category, we develop a score function based on the weighted similarities in this matrix. Extensive experiments demonstrate that our method significantly improves zero-shot performance across various datasets, achieving results that are even comparable to few-shot learning methods.
Authors: Zhong Meng, Zelin Wu, Rohit Prabhavalkar, Cal Peyser, Weiran Wang, Nanxin Chen, Tara N. Sainath, Bhuvana Ramabhadran
Abstract: Neural contextual biasing effectively improves automatic speech recognition (ASR) for crucial phrases within a speaker's context, particularly those that are infrequent in the training data. This work proposes contextual text injection (CTI) to enhance contextual ASR. CTI leverages not only the paired speech-text data, but also a much larger corpus of unpaired text to optimize the ASR model and its biasing component. Unpaired text is converted into speech-like representations and used to guide the model's attention towards relevant bias phrases. Moreover, we introduce a contextual text-injected (CTI) minimum word error rate (MWER) training, which minimizes the expected WER caused by contextual biasing when unpaired text is injected into the model. Experiments show that CTI with 100 billion text sentences can achieve up to 43.3% relative WER reduction from a strong neural biasing model. CTI-MWER provides a further relative improvement of 23.5%.
Authors: Hsuan Su, Hua Farn, Shang-Tse Chen, Hung-yi Lee
Abstract: Recent advancements in large language models (LLMs) have introduced the 'task vector' concept, which has significantly impacted various domains but remains underexplored in speech recognition. This paper presents a novel 'SYN2REAL' task vector for domain adaptation in automatic speech recognition (ASR), specifically targeting text-only domains. Traditional fine-tuning on synthetic speech often results in performance degradation due to acoustic mismatches. To address this issue, we propose creating a 'SYN2REAL' vector by subtracting the parameter differences between models fine-tuned on real and synthetic speech. This vector effectively bridges the gap between the two domains. Experiments on the SLURP dataset demonstrate that our approach yields an average improvement of 11.15% in word error rate for unseen target domains, highlighting the potential of task vectors in enhancing speech domain adaptation.
Authors: Mehdi Jabbari Zideh, Sarika Khushalani Solanki
Abstract: Despite the relentless progress of deep learning models in analyzing the system conditions under cyber-physical events, their abilities are limited in the power system domain due to data availability issues, cost of data acquisition, and lack of interpretation and extrapolation for the data beyond the training windows. In addition, the integration of distributed energy resources (DERs) such as wind and solar generations increases the complexities and nonlinear nature of power systems. Therefore, an interpretable and reliable methodology is of utmost need to increase the confidence of power system operators and their situational awareness for making reliable decisions. This has led to the development of physics-informed neural network (PINN) models as more interpretable, trustworthy, and robust models where the underlying principled laws are integrated into the training process of neural network models to achieve improved performance. This paper proposes a multivariate physics-informed convolutional autoencoder (PIConvAE) model to detect cyber anomalies in power distribution systems with unbalanced configurations and high penetration of DERs. The physical laws are integrated through a customized loss function that embeds the underlying Kirchhoff's circuit laws into the training process of the autoencoder. The performance of the multivariate PIConvAE model is evaluated on two unbalanced power distribution grids, IEEE 123-bus system and a real-world feeder in Riverside, CA. The results show the exceptional performance of the proposed method in detecting various cyber anomalies in both systems. In addition, the model's effectiveness is evaluated in data scarcity scenarios with different training data ratios. Finally, the model's performance is compared with existing machine learning models where the PIConvAE model surpasses other models with considerably higher detection metrics.
Authors: Zihan Ye, Shreyank N. Gowda, Xiaobo Jin, Xiaowei Huang, Haotian Xu, Yaochu Jin, Kaizhu Huang
Abstract: Zero-Shot Learning (ZSL) aims to enable classifiers to identify unseen classes by enhancing data efficiency at the class level. This is achieved by generating image features from pre-defined semantics of unseen classes. However, most current approaches heavily depend on the number of samples from seen classes, i.e. they do not consider instance-level effectiveness. In this paper, we demonstrate that limited seen examples generally result in deteriorated performance of generative models. To overcome these challenges, we propose ZeroDiff, a Diffusion-based Generative ZSL model. This unified framework incorporates diffusion models to improve data efficiency at both the class and instance levels. Specifically, for instance-level effectiveness, ZeroDiff utilizes a forward diffusion chain to transform limited data into an expanded set of noised data. For class-level effectiveness, we design a two-branch generation structure that consists of a Diffusion-based Feature Generator (DFG) and a Diffusion-based Representation Generator (DRG). DFG focuses on learning and sampling the distribution of cross-entropy-based features, whilst DRG learns the supervised contrastive-based representation to boost the zero-shot capabilities of DFG. Additionally, we employ three discriminators to evaluate generated features from various aspects and introduce a Wasserstein-distance-based mutual learning loss to transfer knowledge among discriminators, thereby enhancing guidance for generation. Demonstrated through extensive experiments on three popular ZSL benchmarks, our ZeroDiff not only achieves significant improvements over existing ZSL methods but also maintains robust performance even with scarce training data. Code will be released upon acceptance.
Authors: Yan Huang, Xiang Li, Yipeng Shen, Niao He, Jinming Xu
Abstract: In this paper, we show that applying adaptive methods directly to distributed minimax problems can result in non-convergence due to inconsistency in locally computed adaptive stepsizes. To address this challenge, we propose D-AdaST, a Distributed Adaptive minimax method with Stepsize Tracking. The key strategy is to employ an adaptive stepsize tracking protocol involving the transmission of two extra (scalar) variables. This protocol ensures the consistency among stepsizes of nodes, eliminating the steady-state error due to the lack of coordination of stepsizes among nodes that commonly exists in vanilla distributed adaptive methods, and thus guarantees exact convergence. For nonconvex-strongly-concave distributed minimax problems, we characterize the specific transient times that ensure time-scale separation of stepsizes and quasi-independence of networks, leading to a near-optimal convergence rate of $\tilde{\mathcal{O}} \left( \epsilon ^{-\left( 4+\delta \right)} \right)$ for any small $\delta > 0$, matching that of the centralized counterpart. To our best knowledge, D-AdaST is the first distributed adaptive method achieving near-optimal convergence without knowing any problem-dependent parameters for nonconvex minimax problems. Extensive experiments are conducted to validate our theoretical results.
Authors: Chen Jia
Abstract: Knowledge distillation (KD) has been shown to be highly effective in guiding a student model with a larger teacher model and achieving practical benefits in improving the computational and memory efficiency for large language models (LLMs). State-of-the-art KD methods for LLMs mostly rely on minimizing explicit distribution distance between teacher and student probability predictions. Instead of optimizing these mandatory behaviour cloning objectives, we explore an imitation learning strategy for KD of LLMs. In particular, we minimize the imitation gap by matching the action-value moments of the teacher's behavior from both on- and off-policy perspectives. To achieve this action-value moment-matching goal, we propose an adversarial training algorithm to jointly estimate the moment-matching distance and optimize the student policy to minimize it. Results from both task-agnostic instruction-following experiments and task-specific experiments demonstrate the effectiveness of our method and achieve new state-of-the-art performance.
Authors: Andrea Montanari, Kangjie Zhou
Abstract: Given $d$-dimensional standard Gaussian vectors $\boldsymbol{x}_1,\dots, \boldsymbol{x}_n$, we consider the set of all empirical distributions of its $m$-dimensional projections, for $m$ a fixed constant. Diaconis and Freedman (1984) proved that, if $n/d\to \infty$, all such distributions converge to the standard Gaussian distribution. In contrast, we study the proportional asymptotics, whereby $n,d\to \infty$ with $n/d\to \alpha \in (0, \infty)$. In this case, the projection of the data points along a typical random subspace is again Gaussian, but the set $\mathscr{F}_{m,\alpha}$ of all probability distributions that are asymptotically feasible as $m$-dimensional projections contains non-Gaussian distributions corresponding to exceptional subspaces. Non-rigorous methods from statistical physics yield an indirect characterization of $\mathscr{F}_{m,\alpha}$ in terms of a generalized Parisi formula. Motivated by the goal of putting this formula on a rigorous basis, and to understand whether these projections can be found efficiently, we study the subset $\mathscr{F}^{\rm alg}_{m,\alpha}\subseteq \mathscr{F}_{m,\alpha}$ of distributions that can be realized by a class of iterative algorithms. We prove that this set is characterized by a certain stochastic optimal control problem, and obtain a dual characterization of this problem in terms of a variational principle that extends Parisi's formula. As a byproduct, we obtain computationally achievable values for a class of random optimization problems including `generalized spherical perceptron' models.
Authors: Yexin Zhang, Chenyi Zhang, Cong Fang, Liwei Wang, Tongyang Li
Abstract: Finite-sum optimization has wide applications in machine learning, covering important problems such as support vector machines, regression, etc. In this paper, we initiate the study of solving finite-sum optimization problems by quantum computing. Specifically, let $f_1,\ldots,f_n\colon\mathbb{R}^d\to\mathbb{R}$ be $\ell$-smooth convex functions and $\psi\colon\mathbb{R}^d\to\mathbb{R}$ be a $\mu$-strongly convex proximal function. The goal is to find an $\epsilon$-optimal point for $F(\mathbf{x})=\frac{1}{n}\sum_{i=1}^n f_i(\mathbf{x})+\psi(\mathbf{x})$. We give a quantum algorithm with complexity $\tilde{O}\big(n+\sqrt{d}+\sqrt{\ell/\mu}\big(n^{1/3}d^{1/3}+n^{-2/3}d^{5/6}\big)\big)$, improving the classical tight bound $\tilde{\Theta}\big(n+\sqrt{n\ell/\mu}\big)$. We also prove a quantum lower bound $\tilde{\Omega}(n+n^{3/4}(\ell/\mu)^{1/4})$ when $d$ is large enough. Both our quantum upper and lower bounds can extend to the cases where $\psi$ is not necessarily strongly convex, or each $f_i$ is Lipschitz but not necessarily smooth. In addition, when $F$ is nonconvex, our quantum algorithm can find an $\epsilon$-critial point using $\tilde{O}(n+\ell(d^{1/3}n^{1/3}+\sqrt{d})/\epsilon^2)$ queries.
Authors: Yifei Wang, Dizhan Xue, Shengjie Zhang, Shengsheng Qian
Abstract: With the prosperity of large language models (LLMs), powerful LLM-based intelligent agents have been developed to provide customized services with a set of user-defined tools. State-of-the-art methods for constructing LLM agents adopt trained LLMs and further fine-tune them on data for the agent task. However, we show that such methods are vulnerable to our proposed backdoor attacks named BadAgent on various agent tasks, where a backdoor can be embedded by fine-tuning on the backdoor data. At test time, the attacker can manipulate the deployed LLM agents to execute harmful operations by showing the trigger in the agent input or environment. To our surprise, our proposed attack methods are extremely robust even after fine-tuning on trustworthy data. Though backdoor attacks have been studied extensively in natural language processing, to the best of our knowledge, we could be the first to study them on LLM agents that are more dangerous due to the permission to use external tools. Our work demonstrates the clear risk of constructing LLM agents based on untrusted LLMs or data. Our code is public at https://github.com/DPamK/BadAgent
Authors: Ali Malik, Stephen Mayhew, Chris Piech, Klinton Bicknell
Abstract: We study the problem of controlling the difficulty level of text generated by Large Language Models (LLMs) for contexts where end-users are not fully proficient, such as language learners. Using a novel framework, we evaluate the effectiveness of several key approaches for this task, including few-shot prompting, supervised finetuning, and reinforcement learning (RL), utilising both GPT-4 and open source alternatives like LLama2-7B and Mistral-7B. Our findings reveal a large performance gap between GPT-4 and the open source models when using prompt-based strategies. However, we show how to bridge this gap with a careful combination of finetuning and RL alignment. Our best model, CALM (CEFR-Aligned Language Model), surpasses the performance of GPT-4 and other strategies, at only a fraction of the cost. We further validate the quality of our results through a small-scale human study.
Authors: Sheng Zhou, Yukuan Jia, Ruiqing Mao, Zhaojun Nan, Yuxuan Sun, Zhisheng Niu
Abstract: Collaborative Perception (CP) has shown great potential to achieve more holistic and reliable environmental perception in intelligent unmanned systems (IUSs). However, implementing CP still faces key challenges due to the characteristics of the CP task and the dynamics of wireless channels. In this article, a task-oriented wireless communication framework is proposed to jointly optimize the communication scheme and the CP procedure. We first propose channel-adaptive compression and robust fusion approaches to extract and exploit the most valuable semantic information under wireless communication constraints. We then propose a task-oriented distributed scheduling algorithm to identify the best collaborators for CP under dynamic environments. The main idea is learning while scheduling, where the collaboration utility is effectively learned with low computation and communication overhead. Case studies are carried out in connected autonomous driving scenarios to verify the proposed framework. Finally, we identify several future research directions.
Authors: Samar Agnihotri, Renu Rameshan, Ritwik Ghosal
Abstract: Lossless image compression is required in various applications to reduce storage or transmission costs of images, while requiring the reconstructed images to have zero information loss compared to the original. Existing lossless image compression methods either have simple design but poor compression performance, or complex design, better performance, but with no performance guarantees. In our endeavor to develop a lossless image compression method with low complexity and guaranteed performance, we argue that compressibility of a color image is essentially derived from the patterns in its spatial structure, intensity variations, and color variations. Thus, we divide the overall design of a lossless image compression scheme into three parts that exploit corresponding redundancies. We further argue that the binarized version of an image captures its fundamental spatial structure and in this work, we propose a scheme for lossless compression of binary images. The proposed scheme first learns dictionaries of $16\times16$, $8\times8$, $4\times4$, and $2\times 2$ square pixel patterns from various datasets of binary images. It then uses these dictionaries to encode binary images. These dictionaries have various interesting properties that are further exploited to construct an efficient scheme. Our preliminary results show that the proposed scheme consistently outperforms existing conventional and learning based lossless compression approaches, and provides, on average, as much as $1.5\times$ better performance than a common general purpose lossless compression scheme (WebP), more than $3\times$ better performance than a state of the art learning based scheme, and better performance than a specialized scheme for binary image compression (JBIG2).
Authors: Zhewen Yu, Sudarshan Sreeram, Krish Agrawal, Junyi Wu, Alexander Montgomerie-Corcoran, Cheng Zhang, Jianyi Cheng, Christos-Savvas Bouganis, Yiren Zhao
Abstract: Deep Neural Networks (DNNs) excel in learning hierarchical representations from raw data, such as images, audio, and text. To compute these DNN models with high performance and energy efficiency, these models are usually deployed onto customized hardware accelerators. Among various accelerator designs, dataflow architecture has shown promising performance due to its layer-pipelined structure and its scalability in data parallelism. Exploiting weights and activations sparsity can further enhance memory storage and computation efficiency. However, existing approaches focus on exploiting sparsity in non-dataflow accelerators, which cannot be applied onto dataflow accelerators because of the large hardware design space introduced. As such, this could miss opportunities to find an optimal combination of sparsity features and hardware designs. In this paper, we propose a novel approach to exploit unstructured weights and activations sparsity for dataflow accelerators, using software and hardware co-optimization. We propose a Hardware-Aware Sparsity Search (HASS) to systematically determine an efficient sparsity solution for dataflow accelerators. Over a set of models, we achieve an efficiency improvement ranging from 1.3$\times$ to 4.2$\times$ compared to existing sparse designs, which are either non-dataflow or non-hardware-aware. Particularly, the throughput of MobileNetV3 can be optimized to 4895 images per second. HASS is open-source: \url{https://github.com/Yu-Zhewen/HASS}
Authors: Ryo Fujii, Hideo Saito, Hiroyuki Kajita
Abstract: Surgical tool detection is a fundamental task for understanding egocentric open surgery videos. However, detecting surgical tools presents significant challenges due to their highly imbalanced class distribution, similar shapes and similar textures, and heavy occlusion. The lack of a comprehensive large-scale dataset compounds these challenges. In this paper, we introduce EgoSurgery-Tool, an extension of the existing EgoSurgery-Phase dataset, which contains real open surgery videos captured using an egocentric camera attached to the surgeon's head, along with phase annotations. EgoSurgery-Tool has been densely annotated with surgical tools and comprises over 49K surgical tool bounding boxes across 15 categories, constituting a large-scale surgical tool detection dataset. EgoSurgery-Tool also provides annotations for hand detection with over 46K hand-bounding boxes, capturing hand-object interactions that are crucial for understanding activities in egocentric open surgery. EgoSurgery-Tool is superior to existing datasets due to its larger scale, greater variety of surgical tools, more annotations, and denser scenes. We conduct a comprehensive analysis of EgoSurgery-Tool using nine popular object detectors to assess their effectiveness in both surgical tool and hand detection. The dataset will be released at https://github.com/Fujiry0/EgoSurgery.
Authors: Jacob Bitterman, Daniel Levi, Hilel Hagai Diamandi, Sharon Gannot, Tal Rosenwein
Abstract: This paper focuses on room fingerprinting, a task involving the analysis of an audio recording to determine the specific volume and shape of the room in which it was captured. While it is relatively straightforward to determine the basic room parameters from the Room Impulse Responses (RIR), doing so from a speech signal is a cumbersome task. To address this challenge, we introduce a dual-encoder architecture that facilitates the estimation of room parameters directly from speech utterances. During pre-training, one encoder receives the RIR while the other processes the reverberant speech signal. A contrastive loss function is employed to embed the speech and the acoustic response jointly. In the fine-tuning stage, the specific classification task is trained. In the test phase, only the reverberant utterance is available, and its embedding is used for the task of room shape classification. The proposed scheme is extensively evaluated using simulated acoustic environments.
Authors: Ke Liu, Weian Mao, Shuaike Shen, Xiaoran Jiao, Zheng Sun, Hao Chen, Chunhua Shen
Abstract: Motif scaffolding seeks to design scaffold structures for constructing proteins with functions derived from the desired motif, which is crucial for the design of vaccines and enzymes. Previous works approach the problem by inpainting or conditional generation. Both of them can only scaffold motifs with fixed positions, and the conditional generation cannot guarantee the presence of motifs. However, prior knowledge of the relative motif positions in a protein is not readily available, and constructing a protein with multiple functions in one protein is more general and significant because of the synergies between functions. We propose a Floating Anchor Diffusion (FADiff) model. FADiff allows motifs to float rigidly and independently in the process of diffusion, which guarantees the presence of motifs and automates the motif position design. Our experiments demonstrate the efficacy of FADiff with high success rates and designable novel scaffolds. To the best of our knowledge, FADiff is the first work to tackle the challenge of scaffolding multiple motifs without relying on the expertise of relative motif positions in the protein. Code is available at https://github.com/aim-uofa/FADiff.
Authors: Jianyu Liu, Wei Chen, Yong Zhang, Zhenfeng Chen, Bin Wan, Jinwei Hu
Abstract: In order to solve the problems such as difficult to extract effective features and low accuracy of sales volume prediction caused by complex relationships such as market sales volume in time series prediction, we proposed a time series prediction method of market sales volume based on Sequential General VMD and spatial smoothing Long short-term memory neural network (SS-LSTM) combination model. Firstly, the spatial smoothing algorithm is used to decompose and calculate the sample data of related industry sectors affected by the linkage effect of market sectors, extracting modal features containing information via Sequential General VMD on overall market and specific price trends; Then, according to the background of different Market data sets, LSTM network is used to model and predict the price of fundamental data and modal characteristics. The experimental results of data prediction with seasonal and periodic trends show that this method can achieve higher price prediction accuracy and more accurate accuracy in specific market contexts compared to traditional prediction methods Describe the changes in market sales volume.
Authors: Erik Landolsi, Fredrik Kahl
Abstract: Few-shot image classification involves classifying images using very few training examples. Recent vision foundation models show excellent few-shot transfer abilities, but are large and slow at inference. Using knowledge distillation, the capabilities of high-performing but slow models can be transferred to tiny, efficient models. However, common distillation methods require a large set of unlabeled data, which is not available in the few-shot setting. To overcome this lack of data, there has been a recent interest in using synthetic data. We expand on this work by presenting a novel diffusion model inversion technique (TINT) combining the diversity of textual inversion with the specificity of null-text inversion. Using this method in a few-shot distillation pipeline leads to state-of-the-art accuracy among small student models on popular benchmarks, while being significantly faster than prior work. This allows us to push even tiny models to high accuracy using only a tiny application-specific dataset, albeit relying on extra data for pre-training. Popular few-shot benchmarks involve evaluation over a large number of episodes, which is computationally cumbersome for methods involving synthetic data generation. Therefore, we also present a theoretical analysis on how the variance of the accuracy estimator depends on the number of episodes and query examples, and use these results to lower the computational effort required for method evaluation. In addition, to further motivate the use of generative models in few-shot distillation, we demonstrate that our method performs better compared to training on real data mined from the dataset used to train the diffusion model. Source code will be made available at https://github.com/pixwse/tiny2.
Authors: Hao Li, Yuping Wu, Viktor Schlegel, Riza Batista-Navarro, Tharindu Madusanka, Iqra Zahid, Jiayan Zeng, Xiaochi Wang, Xinran He, Yizhi Li, Goran Nenadic
Abstract: With the recent advances of large language models (LLMs), it is no longer infeasible to build an automated debate system that helps people to synthesise persuasive arguments. Previous work attempted this task by integrating multiple components. In our work, we introduce an argument mining dataset that captures the end-to-end process of preparing an argumentative essay for a debate, which covers the tasks of claim and evidence identification (Task 1 ED), evidence convincingness ranking (Task 2 ECR), argumentative essay summarisation and human preference ranking (Task 3 ASR) and metric learning for automated evaluation of resulting essays, based on human feedback along argument quality dimensions (Task 4 SQE). Our dataset contains 14k examples of claims that are fully annotated with the various properties supporting the aforementioned tasks. We evaluate multiple generative baselines for each of these tasks, including representative LLMs. We find, that while they show promising results on individual tasks in our benchmark, their end-to-end performance on all four tasks in succession deteriorates significantly, both in automated measures as well as in human-centred evaluation. This challenge presented by our proposed dataset motivates future research on end-to-end argument mining and summarisation. The repository of this project is available at https://github.com/HarrywillDr/ArgSum-Datatset
Authors: Steinar Laenen, He Sun
Abstract: This paper studies clustering algorithms for dynamically evolving graphs $\{G_t\}$, in which new edges (and potential new vertices) are added into a graph, and the underlying cluster structure of the graph can gradually change. The paper proves that, under some mild condition on the cluster-structure, the clusters of the final graph $G_T$ of $n_T$ vertices at time $T$ can be well approximated by a dynamic variant of the spectral clustering algorithm. The algorithm runs in amortised update time $O(1)$ and query time $o(n_T)$. Experimental studies on both synthetic and real-world datasets further confirm the practicality of our designed algorithm.
Authors: Wei Chen, Yuanyuan Yang, Jianyu Liu
Abstract: Accurate prediction of financial time series is a key concern for market economy makers and investors. The article selects online store sales and Australian beer sales as representatives of non-stationary, trending, and seasonal financial time series, and constructs a new SGVMD-ARIMA combination model in a non-linear combination way to predict financial time series. The ARIMA model, LSTM model, and other classic decomposition prediction models are used as control models to compare the accuracy of different models. The empirical results indicate that the constructed combination prediction model has universal advantages over the single prediction model and linear combination prediction model of the control group. Within the prediction interval, our proposed combination model has improved advantages over traditional decomposition prediction control group models.
Authors: Yihang Chen, Fanghui Liu, Taiji Suzuki, Volkan Cevher
Abstract: This paper studies kernel ridge regression in high dimensions under covariate shifts and analyzes the role of importance re-weighting. We first derive the asymptotic expansion of high dimensional kernels under covariate shifts. By a bias-variance decomposition, we theoretically demonstrate that the re-weighting strategy allows for decreasing the variance. For bias, we analyze the regularization of the arbitrary or well-chosen scale, showing that the bias can behave very differently under different regularization scales. In our analysis, the bias and variance can be characterized by the spectral decay of a data-dependent regularized kernel: the original kernel matrix associated with an additional re-weighting matrix, and thus the re-weighting strategy can be regarded as a data-dependent regularization for better understanding. Besides, our analysis provides asymptotic expansion of kernel functions/vectors under covariate shift, which has its own interest.
Authors: Jiate Li, Meng Pang, Yun Dong, Jinyuan Jia, Binghui Wang
Abstract: Explainable Graph Neural Network (GNN) has emerged recently to foster the trust of using GNNs. Existing GNN explainers are developed from various perspectives to enhance the explanation performance. We take the first step to study GNN explainers under adversarial attack--We found that an adversary slightly perturbing graph structure can ensure GNN model makes correct predictions, but the GNN explainer yields a drastically different explanation on the perturbed graph. Specifically, we first formulate the attack problem under a practical threat model (i.e., the adversary has limited knowledge about the GNN explainer and a restricted perturbation budget). We then design two methods (i.e., one is loss-based and the other is deduction-based) to realize the attack. We evaluate our attacks on various GNN explainers and the results show these explainers are fragile.
Authors: Jacy Anthis, Kristian Lum, Michael Ekstrand, Avi Feller, Alexander D'Amour, Chenhao Tan
Abstract: The need for fair AI is increasingly clear in the era of general-purpose systems such as ChatGPT, Gemini, and other large language models (LLMs). However, the increasing complexity of human-AI interaction and its social impacts have raised questions of how fairness standards could be applied. Here, we review the technical frameworks that machine learning researchers have used to evaluate fairness, such as group fairness and fair representations, and find that their application to LLMs faces inherent limitations. We show that each framework either does not logically extend to LLMs or presents a notion of fairness that is intractable for LLMs, primarily due to the multitudes of populations affected, sensitive attributes, and use cases. To address these challenges, we develop guidelines for the more realistic goal of achieving fairness in particular use cases: the criticality of context, the responsibility of LLM developers, and the need for stakeholder participation in an iterative process of design and evaluation. Moreover, it may eventually be possible and even necessary to use the general-purpose capabilities of AI systems to address fairness challenges as a form of scalable AI-assisted alignment.
Authors: Ziyun Cui, Ziyang Zhang, Wen Wu, Guangzhi Sun, Chao Zhang
Abstract: Advances in large language models raise the question of how alignment techniques will adapt as models become increasingly complex and humans will only be able to supervise them weakly. Weak-to-Strong mimics such a scenario where weak model supervision attempts to harness the full capabilities of a much stronger model. This work extends Weak-to-Strong to WeakS-to-Strong by exploring an ensemble of weak models which simulate the variability in human opinions. Confidence scores are estimated using a Bayesian approach to guide the WeakS-to-Strong generalization. Furthermore, we extend the application of WeakS-to-Strong from text classification tasks to text generation tasks where more advanced strategies are investigated for supervision. Moreover, direct preference optimization is applied to advance the student model's preference learning, beyond the basic learning framework of teacher forcing. Results demonstrate the effectiveness of the proposed approach for the reliability of a strong student model, showing potential for superalignment.
Authors: Qutub Syed Sha, Michael Paulitsch, Karthik Pattabiraman, Korbinian Hagn, Fabian Oboril, Cornelius Buerkle, Kay-Ulrich Scholl, Gereon Hinz, Alois Knoll
Abstract: As transformer-based object detection models progress, their impact in critical sectors like autonomous vehicles and aviation is expected to grow. Soft errors causing bit flips during inference have significantly impacted DNN performance, altering predictions. Traditional range restriction solutions for CNNs fall short for transformers. This study introduces the Global Clipper and Global Hybrid Clipper, effective mitigation strategies specifically designed for transformer-based models. It significantly enhances their resilience to soft errors and reduces faulty inferences to ~ 0\%. We also detail extensive testing across over 64 scenarios involving two transformer models (DINO-DETR and Lite-DETR) and two CNN models (YOLOv3 and SSD) using three datasets, totalling approximately 3.3 million inferences, to assess model robustness comprehensively. Moreover, the paper explores unique aspects of attention blocks in transformers and their operational differences from CNNs.
Authors: Amelia Kawasaki, Andrew Davis, Houssam Abbas
Abstract: The widespread adoption of Large Language Models (LLMs), exemplified by OpenAI's ChatGPT, brings to the forefront the imperative to defend against adversarial threats on these models. These attacks, which manipulate an LLM's output by introducing malicious inputs, undermine the model's integrity and the trust users place in its outputs. In response to this challenge, our paper presents an innovative defensive strategy, given white box access to an LLM, that harnesses residual activation analysis between transformer layers of the LLM. We apply an established methodology for analyzing distinctive activation patterns in the residual streams for a novel result of attack prompt classification. We curate multiple datasets to demonstrate how this method of classification has high accuracy across multiple types of attack scenarios, including our newly-created attack dataset. Furthermore, we enhance the model's resilience by integrating safety fine-tuning techniques for LLMs in order to measure its effect on our capability to detect attacks. The results underscore the effectiveness of our approach in enhancing the detection and mitigation of adversarial inputs, advancing the security framework within which LLMs operate.
Authors: Michael Eichelbeck, Hannah Markgraf, Matthias Althoff
Abstract: The growing complexity of power system management has led to an increased interest in the use of reinforcement learning (RL). However, no tool for comprehensive and realistic benchmarking of RL in smart grids exists. One prerequisite for such a comparison is a safeguarding mechanism since vanilla RL controllers can not guarantee the satisfaction of system constraints. Other central requirements include flexible modeling of benchmarking scenarios, credible baselines, and the possibility to investigate the impact of forecast uncertainties. Our Python tool CommonPower is the first modular framework addressing these needs. CommonPower offers a unified interface for single-agent and multi-agent RL training algorithms and includes a built-in model predictive control approach based on a symbolic representation of the system equations. This makes it possible to combine model predictive controllers with RL controllers in the same system. Leveraging the symbolic system model, CommonPower facilitates the study of safeguarding strategies via the flexible formulation of safety layers. Furthermore equipped with a generic forecasting interface, CommonPower constitutes a versatile tool significantly augmenting the exploration of safe RL controllers in smart grids on several dimensions.
Authors: Thomas Pouplin, Alan Jeffares, Nabeel Seedat, Mihaela van der Schaar
Abstract: Constructing valid prediction intervals rather than point estimates is a well-established approach for uncertainty quantification in the regression setting. Models equipped with this capacity output an interval of values in which the ground truth target will fall with some prespecified probability. This is an essential requirement in many real-world applications where simple point predictions' inability to convey the magnitude and frequency of errors renders them insufficient for high-stakes decisions. Quantile regression is a leading approach for obtaining such intervals via the empirical estimation of quantiles in the (non-parametric) distribution of outputs. This method is simple, computationally inexpensive, interpretable, assumption-free, and effective. However, it does require that the specific quantiles being learned are chosen a priori. This results in (a) intervals that are arbitrarily symmetric around the median which is sub-optimal for realistic skewed distributions, or (b) learning an excessive number of intervals. In this work, we propose Relaxed Quantile Regression (RQR), a direct alternative to quantile regression based interval construction that removes this arbitrary constraint whilst maintaining its strengths. We demonstrate that this added flexibility results in intervals with an improvement in desirable qualities (e.g. mean width) whilst retaining the essential coverage guarantees of quantile regression.
Authors: Federico Bassetti, Marco Gherardi, Alessandro Ingrosso, Mauro Pastore, Pietro Rotondo
Abstract: Deep linear networks have been extensively studied, as they provide simplified models of deep learning. However, little is known in the case of finite-width architectures with multiple outputs and convolutional layers. In this manuscript, we provide rigorous results for the statistics of functions implemented by the aforementioned class of networks, thus moving closer to a complete characterization of feature learning in the Bayesian setting. Our results include: (i) an exact and elementary non-asymptotic integral representation for the joint prior distribution over the outputs, given in terms of a mixture of Gaussians; (ii) an analytical formula for the posterior distribution in the case of squared error loss function (Gaussian likelihood); (iii) a quantitative description of the feature learning infinite-width regime, using large deviation theory. From a physical perspective, deep architectures with multiple outputs or convolutional layers represent different manifestations of kernel shape renormalization, and our work provides a dictionary that translates this physics intuition and terminology into rigorous Bayesian statistics.
Authors: Arpan Losalka, Jonathan Scarlett
Abstract: We consider the problem of sequentially maximizing an unknown function $f$ over a set of actions of the form $(s,\mathbf{x})$, where the selected actions must satisfy a safety constraint with respect to an unknown safety function $g$. We model $f$ and $g$ as lying in a reproducing kernel Hilbert space (RKHS), which facilitates the use of Gaussian process methods. While existing works for this setting have provided algorithms that are guaranteed to identify a near-optimal safe action, the problem of attaining low cumulative regret has remained largely unexplored, with a key challenge being that expanding the safe region can incur high regret. To address this challenge, we show that if $g$ is monotone with respect to just the single variable $s$ (with no such constraint on $f$), sublinear regret becomes achievable with our proposed algorithm. In addition, we show that a modified version of our algorithm is able to attain sublinear regret (for suitably defined notions of regret) for the task of finding a near-optimal $s$ corresponding to every $\mathbf{x}$, as opposed to only finding the global safe optimum. Our findings are supported with empirical evaluations on various objective and safety functions.
Authors: Ohad Cohen, Gershon Hazan, Sharon Gannot
Abstract: Most emotion recognition systems fail in real-life situations (in the wild scenarios) where the audio is contaminated by reverberation. Our study explores new methods to alleviate the performance degradation of Speech Emotion Recognition (SER) algorithms and develop a more robust system for adverse conditions. We propose processing multi-microphone signals to address these challenges and improve emotion classification accuracy. We adopt a state-of-the-art transformer model, the Hierarchical Token-semantic Audio Transformer (HTS-AT), to handle multi-channel audio inputs. We evaluate two strategies: averaging mel-spectrograms across channels and summing patch-embedded representations. Our multimicrophone model achieves superior performance compared to single-channel baselines when tested on real-world reverberant environments.
Authors: Xingrun Xing, Zheng Zhang, Ziyi Ni, Shitao Xiao, Yiming Ju, Siqi Fan, Yequan Wang, Jiajun Zhang, Guoqi Li
Abstract: Towards energy-efficient artificial intelligence similar to the human brain, the bio-inspired spiking neural networks (SNNs) have advantages of biological plausibility, event-driven sparsity, and binary activation. Recently, large-scale language models exhibit promising generalization capability, making it a valuable issue to explore more general spike-driven models. However, the binary spikes in existing SNNs fail to encode adequate semantic information, placing technological challenges for generalization. This work proposes the first fully spiking mechanism for general language tasks, including both discriminative and generative ones. Different from previous spikes with {0,1} levels, we propose a more general spike formulation with bi-directional, elastic amplitude, and elastic frequency encoding, while still maintaining the addition nature of SNNs. In a single time step, the spike is enhanced by direction and amplitude information; in spike frequency, a strategy to control spike firing rate is well designed. We plug this elastic bi-spiking mechanism in language modeling, named SpikeLM. It is the first time to handle general language tasks with fully spike-driven models, which achieve much higher accuracy than previously possible. SpikeLM also greatly bridges the performance gap between SNNs and ANNs in language modeling. Our code is available at https://github.com/Xingrun-Xing/SpikeLM.
Authors: Isma\"el Castillo, Paul Egels
Abstract: We consider deep neural networks in a Bayesian framework with a prior distribution sampling the network weights at random. Following a recent idea of Agapiou and Castillo (2023), who show that heavy-tailed prior distributions achieve automatic adaptation to smoothness, we introduce a simple Bayesian deep learning prior based on heavy-tailed weights and ReLU activation. We show that the corresponding posterior distribution achieves near-optimal minimax contraction rates, simultaneously adaptive to both intrinsic dimension and smoothness of the underlying function, in a variety of contexts including nonparametric regression, geometric data and Besov spaces. While most works so far need a form of model selection built-in within the prior distribution, a key aspect of our approach is that it does not require to sample hyperparameters to learn the architecture of the network. We also provide variational Bayes counterparts of the results, that show that mean-field variational approximations still benefit from near-optimal theoretical support.
Authors: Ali Momeni, Babak Rahmani, Benjamin Scellier, Logan G. Wright, Peter L. McMahon, Clara C. Wanjura, Yuhang Li, Anas Skalli, Natalia G. Berloff, Tatsuhiro Onodera, Ilker Oguz, Francesco Morichetti, Philipp del Hougne, Manuel Le Gallo, Abu Sebastian, Azalia Mirhoseini, Cheng Zhang, Danijela Markovi\'c, Daniel Brunner, Christophe Moser, Sylvain Gigan, Florian Marquardt, Aydogan Ozcan, Julie Grollier, Andrea J. Liu, Demetri Psaltis, Andrea Al\`u, Romain Fleury
Abstract: Physical neural networks (PNNs) are a class of neural-like networks that leverage the properties of physical systems to perform computation. While PNNs are so far a niche research area with small-scale laboratory demonstrations, they are arguably one of the most underappreciated important opportunities in modern AI. Could we train AI models 1000x larger than current ones? Could we do this and also have them perform inference locally and privately on edge devices, such as smartphones or sensors? Research over the past few years has shown that the answer to all these questions is likely "yes, with enough research": PNNs could one day radically change what is possible and practical for AI systems. To do this will however require rethinking both how AI models work, and how they are trained - primarily by considering the problems through the constraints of the underlying hardware physics. To train PNNs at large scale, many methods including backpropagation-based and backpropagation-free approaches are now being explored. These methods have various trade-offs, and so far no method has been shown to scale to the same scale and performance as the backpropagation algorithm widely used in deep learning today. However, this is rapidly changing, and a diverse ecosystem of training techniques provides clues for how PNNs may one day be utilized to create both more efficient realizations of current-scale AI models, and to enable unprecedented-scale models.
Authors: Evan Becker, Stefano Soatto
Abstract: In many high-risk machine learning applications it is essential for a model to indicate when it is uncertain about a prediction. While large language models (LLMs) can reach and even surpass human-level accuracy on a variety of benchmarks, their overconfidence in incorrect responses is still a well-documented failure mode. Traditional methods for ML uncertainty quantification can be difficult to directly adapt to LLMs due to the computational cost of implementation and closed-source nature of many models. A variety of black-box methods have recently been proposed, but these often rely on heuristics such as self-verbalized confidence. We instead propose a framework for measuring an LLM's uncertainty with respect to the distribution of generated explanations for an answer. While utilizing explanations is not a new idea in and of itself, by interpreting each possible model+explanation pair as a test-time classifier we can calculate a posterior answer distribution over the most likely of these classifiers. We demonstrate how a specific instance of this framework using explanation entailment as our classifier likelihood improves confidence score metrics (in particular AURC and AUROC) over baselines across five different datasets. We believe these results indicate that our framework is both a well-principled and effective way of quantifying uncertainty in LLMs.
Authors: Mona Ahmadian, Frank Guerin, Andrew Gilbert
Abstract: This paper demonstrates a self-supervised approach for learning semantic video representations. Recent vision studies show that a masking strategy for vision and natural language supervision has contributed to developing transferable visual pretraining. Our goal is to achieve a more semantic video representation by leveraging the text related to the video content during the pretraining in a fully self-supervised manner. To this end, we present FILS, a novel self-supervised video Feature prediction In semantic Language Space (FILS). The vision model can capture valuable structured information by correctly predicting masked feature semantics in language space. It is learned using a patch-wise video-text contrastive strategy, in which the text representations act as prototypes for transforming vision features into a language space, which are then used as targets for semantically meaningful feature prediction using our masked encoder-decoder structure. FILS demonstrates remarkable transferability on downstream action recognition tasks, achieving state-of-the-art on challenging egocentric datasets, like Epic-Kitchens, Something-SomethingV2, Charades-Ego, and EGTEA, using ViT-Base. Our efficient method requires less computation and smaller batches compared to previous works.
Authors: Danilo de Oliveira, Simon Welker, Julius Richter, Timo Gerkmann
Abstract: To obtain improved speech enhancement models, researchers often focus on increasing performance according to specific instrumental metrics. However, when the same metric is used in a loss function to optimize models, it may be detrimental to aspects that the given metric does not see. The goal of this paper is to illustrate the risk of overfitting a speech enhancement model to the metric used for evaluation. For this, we introduce enhancement models that exploit the widely used PESQ measure. Our "PESQetarian" model achieves 3.82 PESQ on VB-DMD while scoring very poorly in a listening experiment. While the obtained PESQ value of 3.82 would imply "state-of-the-art" PESQ-performance on the VB-DMD benchmark, our examples show that when optimizing w.r.t. a metric, an isolated evaluation on the same metric may be misleading. Instead, other metrics should be included in the evaluation and the resulting performance predictions should be confirmed by listening.
Authors: Sonia Bbouzidi, Ghazala Hcini, Imen Jdey, Fadoua Drira
Abstract: Our review explores the comparative analysis between Convolutional Neural Networks (CNNs) and Vision Transformers (ViTs) in the domain of image classification, with a particular focus on clothing classification within the e-commerce sector. Utilizing the Fashion MNIST dataset, we delve into the unique attributes of CNNs and ViTs. While CNNs have long been the cornerstone of image classification, ViTs introduce an innovative self-attention mechanism enabling nuanced weighting of different input data components. Historically, transformers have primarily been associated with Natural Language Processing (NLP) tasks. Through a comprehensive examination of existing literature, our aim is to unveil the distinctions between ViTs and CNNs in the context of image classification. Our analysis meticulously scrutinizes state-of-the-art methodologies employing both architectures, striving to identify the factors influencing their performance. These factors encompass dataset characteristics, image dimensions, the number of target classes, hardware infrastructure, and the specific architectures along with their respective top results. Our key goal is to determine the most appropriate architecture between ViT and CNN for classifying images in the Fashion MNIST dataset within the e-commerce industry, while taking into account specific conditions and needs. We highlight the importance of combining these two architectures with different forms to enhance overall performance. By uniting these architectures, we can take advantage of their unique strengths, which may lead to more precise and reliable models for e-commerce applications. CNNs are skilled at recognizing local patterns, while ViTs are effective at grasping overall context, making their combination a promising strategy for boosting image classification performance.
Authors: Yi-Kai Zhang, Shiyin Lu, Yang Li, Yanqing Ma, Qing-Guo Chen, Zhao Xu, Weihua Luo, Kaifu Zhang, De-Chuan Zhan, Han-Jia Ye
Abstract: Multimodal large language models (MLLMs), initiated with a trained LLM, first align images with text and then fine-tune on multimodal mixed inputs. However, the MLLM catastrophically forgets the text-only instructions, which do not include images and can be addressed within the initial LLM. In this paper, we present Wings, a novel MLLM that excels in both text-only dialogues and multimodal comprehension. Analyzing MLLM attention in multimodal instructions reveals that text-only forgetting is related to the attention shifts from pre-image to post-image text. From that, we construct extra modules that act as the boosted learner to compensate for the attention shift. The complementary visual and textual learners, like "wings" on either side, are connected in parallel within each layer's attention block. Initially, image and text inputs are aligned with visual learners operating alongside the main attention, balancing focus on visual elements. Textual learners are later collaboratively integrated with attention-based routing to blend the outputs of the visual and textual learners. We design the Low-Rank Residual Attention (LoRRA) to guarantee high efficiency for learners. Our experimental results demonstrate that Wings outperforms equally-scaled MLLMs in both text-only and visual question-answering tasks. On a newly constructed Interleaved Image-Text (IIT) benchmark, Wings exhibits superior performance from text-only-rich to multimodal-rich question-answering tasks.
Authors: George H. Chen, Linhong Li, Ren Zuo, Amanda Coston, Jeremy C. Weiss
Abstract: We present a neural network framework for learning a survival model to predict a time-to-event outcome while simultaneously learning a topic model that reveals feature relationships. In particular, we model each subject as a distribution over "topics", where a topic could, for instance, correspond to an age group, a disorder, or a disease. The presence of a topic in a subject means that specific clinical features are more likely to appear for the subject. Topics encode information about related features and are learned in a supervised manner to predict a time-to-event outcome. Our framework supports combining many different topic and survival models; training the resulting joint survival-topic model readily scales to large datasets using standard neural net optimizers with minibatch gradient descent. For example, a special case is to combine LDA with a Cox model, in which case a subject's distribution over topics serves as the input feature vector to the Cox model. We explain how to address practical implementation issues that arise when applying these neural survival-supervised topic models to clinical data, including how to visualize results to assist clinical interpretation. We study the effectiveness of our proposed framework on seven clinical datasets on predicting time until death as well as hospital ICU length of stay, where we find that neural survival-supervised topic models achieve competitive accuracy with existing approaches while yielding interpretable clinical topics that explain feature relationships. Our code is available at: https://github.com/georgehc/survival-topics
Authors: Konstantin Avrachenkov, Maximilien Dreveton
Abstract: Graph-based semi-supervised learning methods combine the graph structure and labeled data to classify unlabeled data. In this work, we study the effect of a noisy oracle on classification. In particular, we derive the Maximum A Posteriori (MAP) estimator for clustering a Degree Corrected Stochastic Block Model (DC-SBM) when a noisy oracle reveals a fraction of the labels. We then propose an algorithm derived from a continuous relaxation of the MAP, and we establish its consistency. Numerical experiments show that our approach achieves promising performance on synthetic and real data sets, even in the case of very noisy labeled data.
Authors: Pablo Villalobos, Anson Ho, Jaime Sevilla, Tamay Besiroglu, Lennart Heim, Marius Hobbhahn
Abstract: We investigate the potential constraints on LLM scaling posed by the availability of public human-generated text data. We forecast the growing demand for training data based on current trends and estimate the total stock of public human text data. Our findings indicate that if current LLM development trends continue, models will be trained on datasets roughly equal in size to the available stock of public human text data between 2026 and 2032, or slightly earlier if models are overtrained. We explore how progress in language modeling can continue when human-generated text datasets cannot be scaled any further. We argue that synthetic data generation, transfer learning from data-rich domains, and data efficiency improvements might support further progress.
Authors: Shachi Deshpande, Charles Marx, Volodymyr Kuleshov
Abstract: We are interested in probabilistic prediction in online settings in which data does not follow a probability distribution. Our work seeks to achieve two goals: (1) producing valid probabilities that accurately reflect model confidence; and (2) ensuring that traditional notions of performance (e.g., high accuracy) still hold. We introduce online algorithms guaranteed to achieve these goals on arbitrary streams of data points, including data chosen by an adversary. Specifically, our algorithms produce forecasts that are (1) calibrated -- i.e., an 80% confidence interval contains the true outcome 80% of the time -- and (2) have low regret relative to a user-specified baseline model. We implement a post-hoc recalibration strategy that provably achieves these goals in regression; previous algorithms applied to classification or achieved (1) but not (2). In the context of Bayesian optimization, an online model-based decision-making task in which the data distribution shifts over time, our method yields accelerated convergence to improved optima.
Authors: Alvin Heng, Abdul Fatir Ansari, Harold Soh
Abstract: We present Flow-Guided Density Ratio Learning (FDRL), a simple and scalable approach to generative modeling which builds on the stale (time-independent) approximation of the gradient flow of entropy-regularized f-divergences introduced in recent work. Specifically, the intractable time-dependent density ratio is approximated by a stale estimator given by a GAN discriminator. This is sufficient in the case of sample refinement, where the source and target distributions of the flow are close to each other. However, this assumption is invalid for generation and a naive application of the stale estimator fails due to the large chasm between the two distributions. FDRL proposes to train a density ratio estimator such that it learns from progressively improving samples during the training process. We show that this simple method alleviates the density chasm problem, allowing FDRL to generate images of dimensions as high as $128\times128$, as well as outperform existing gradient flow baselines on quantitative benchmarks. We also show the flexibility of FDRL with two use cases. First, unconditional FDRL can be easily composed with external classifiers to perform class-conditional generation. Second, FDRL can be directly applied to unpaired image-to-image translation with no modifications needed to the framework. Our code is publicly available at ttps://github.com/clear-nus/fdrl.
Authors: Micah Goldblum, Marc Finzi, Keefer Rowan, Andrew Gordon Wilson
Abstract: No free lunch theorems for supervised learning state that no learner can solve all problems or that all learners achieve exactly the same accuracy on average over a uniform distribution on learning problems. Accordingly, these theorems are often referenced in support of the notion that individual problems require specially tailored inductive biases. While virtually all uniformly sampled datasets have high complexity, real-world problems disproportionately generate low-complexity data, and we argue that neural network models share this same preference, formalized using Kolmogorov complexity. Notably, we show that architectures designed for a particular domain, such as computer vision, can compress datasets on a variety of seemingly unrelated domains. Our experiments show that pre-trained and even randomly initialized language models prefer to generate low-complexity sequences. Whereas no free lunch theorems seemingly indicate that individual problems require specialized learners, we explain how tasks that often require human intervention such as picking an appropriately sized model when labeled data is scarce or plentiful can be automated into a single learning algorithm. These observations justify the trend in deep learning of unifying seemingly disparate problems with an increasingly small set of machine learning models.
Authors: Fabian Schaipp, Ruben Ohana, Michael Eickenberg, Aaron Defazio, Robert M. Gower
Abstract: Training a modern machine learning architecture on a new task requires extensive learning-rate tuning, which comes at a high computational cost. Here we develop new Polyak-type adaptive learning rates that can be used on top of any momentum method, and require less tuning to perform well. We first develop MoMo, a Momentum Model based adaptive learning rate for SGD-M (stochastic gradient descent with momentum). MoMo uses momentum estimates of the losses and gradients sampled at each iteration to build a model of the loss function. Our model makes use of any known lower bound of the loss function by using truncation, e.g. most losses are lower-bounded by zero. The model is then approximately minimized at each iteration to compute the next step. We show how MoMo can be used in combination with any momentum-based method, and showcase this by developing MoMo-Adam, which is Adam with our new model-based adaptive learning rate. We show that MoMo attains a $\mathcal{O}(1/\sqrt{K})$ convergence rate for convex problems with interpolation, needing knowledge of no problem-specific quantities other than the optimal value. Additionally, for losses with unknown lower bounds, we develop on-the-fly estimates of a lower bound, that are incorporated in our model. We show that MoMo and MoMo-Adam improve over SGD-M and Adam in terms of robustness to hyperparameter tuning for training image classifiers on MNIST, CIFAR, and Imagenet, for recommender systems on Criteo, for a transformer model on the translation task IWSLT14, and for a diffusion model.
Authors: Fabrizio Angiulli, Fabio Fassetti, Luca Ferragina
Abstract: Reconstruction error-based neural architectures constitute a classical deep learning approach to anomaly detection which has shown great performances. It consists in training an Autoencoder to reconstruct a set of examples deemed to represent the normality and then to point out as anomalies those data that show a sufficiently large reconstruction error. Unfortunately, these architectures often become able to well reconstruct also the anomalies in the data. This phenomenon is more evident when there are anomalies in the training set. In particular when these anomalies are labeled, a setting called semi-supervised, the best way to train Autoencoders is to ignore anomalies and minimize the reconstruction error on normal data. The goal of this work is to investigate approaches to allow reconstruction error-based architectures to instruct the model to put known anomalies outside of the domain description of the normal data. Specifically, our strategy exploits a limited number of anomalous examples to increase the contrast between the reconstruction error associated with normal examples and those associated with both known and unknown anomalies, thus enhancing anomaly detection performances. The experiments show that this new procedure achieves better performances than the standard Autoencoder approach and the main deep learning techniques for semi-supervised anomaly detection.
Authors: Francesco Pedrotti, Jan Maas, Marco Mondelli
Abstract: Score-based generative models (SGMs) are powerful tools to sample from complex data distributions. Their underlying idea is to (i) run a forward process for time $T_1$ by adding noise to the data, (ii) estimate its score function, and (iii) use such estimate to run a reverse process. As the reverse process is initialized with the stationary distribution of the forward one, the existing analysis paradigm requires $T_1\to\infty$. This is however problematic: from a theoretical viewpoint, for a given precision of the score approximation, the convergence guarantee fails as $T_1$ diverges; from a practical viewpoint, a large $T_1$ increases computational costs and leads to error propagation. This paper addresses the issue by considering a version of the popular predictor-corrector scheme: after running the forward process, we first estimate the final distribution via an inexact Langevin dynamics and then revert the process. Our key technical contribution is to provide convergence guarantees which require to run the forward process only for a fixed finite time $T_1$. Our bounds exhibit a mild logarithmic dependence on the input dimension and the subgaussian norm of the target distribution, have minimal assumptions on the data, and require only to control the $L^2$ loss on the score approximation, which is the quantity minimized in practice.
Authors: Kha-Dinh Luong, Mert Kosan, Arlei Lopes Da Silva, Ambuj Singh
Abstract: Explaining the decisions made by machine learning models for high-stakes applications is critical for increasing transparency and guiding improvements to these decisions. This is particularly true in the case of models for graphs, where decisions often depend on complex patterns combining rich structural and attribute data. While recent work has focused on designing so-called post-hoc explainers, the broader question of what constitutes a good explanation remains open. One intuitive property is that explanations should be sufficiently informative to reproduce the predictions given the data. In other words, a good explainer can be repurposed as a predictor. Post-hoc explainers do not achieve this goal as their explanations are highly dependent on fixed model parameters (e.g., learned GNN weights). To address this challenge, we propose RAGE (Robust Ante-hoc Graph Explainer), a novel and flexible ante-hoc explainer designed to discover explanations for graph neural networks using bilevel optimization, with a focus on the chemical domain. RAGE can effectively identify molecular substructures that contain the full information needed for prediction while enabling users to rank these explanations in terms of relevance. Our experiments on various molecular classification tasks show that RAGE explanations are better than existing post-hoc and ante-hoc approaches.
Authors: Elena Orlova, Aleksei Ustimenko, Ruoxi Jiang, Peter Y. Lu, Rebecca Willett
Abstract: This paper introduces a novel deep-learning-based approach for numerical simulation of a time-evolving Schr\"odinger equation inspired by stochastic mechanics and generative diffusion models. Unlike existing approaches, which exhibit computational complexity that scales exponentially in the problem dimension, our method allows us to adapt to the latent low-dimensional structure of the wave function by sampling from the Markovian diffusion. Depending on the latent dimension, our method may have far lower computational complexity in higher dimensions. Moreover, we propose novel equations for stochastic quantum mechanics, resulting in quadratic computational complexity with respect to the number of dimensions. Numerical simulations verify our theoretical findings and show a significant advantage of our method compared to other deep-learning-based approaches used for quantum mechanics.
Authors: Minting Pan, Yitao Zheng, Yunbo Wang, Xiaokang Yang
Abstract: Pretraining reinforcement learning (RL) models on offline datasets is a promising way to improve their training efficiency in online tasks, but challenging due to the inherent mismatch in dynamics and behaviors across various tasks. We present a model-based RL method that learns to transfer potentially useful dynamics and action demonstrations from offline data to a novel task. The main idea is to use the world models not only as simulators for behavior learning but also as tools to measure the task relevance for both dynamics representation transfer and policy transfer. We build a time-varying, domain-selective distillation loss to generate a set of offline-to-online similarity weights. These weights serve two purposes: (i) adaptively transferring the task-agnostic knowledge of physical dynamics to facilitate world model training, and (ii) learning to replay relevant source actions to guide the target policy. We demonstrate the advantages of our approach compared with the state-of-the-art methods in Meta-World and DeepMind Control Suite.
Authors: Runze Liu, Yali Du, Fengshuo Bai, Jiafei Lyu, Xiu Li
Abstract: In preference-based Reinforcement Learning (RL), obtaining a large number of preference labels are both time-consuming and costly. Furthermore, the queried human preferences cannot be utilized for the new tasks. In this paper, we propose Zero-shot Cross-task Preference Alignment and Robust Reward Learning (PEARL), which learns policies from cross-task preference transfer without any human labels of the target task. Our contributions include two novel components that facilitate the transfer and learning process. The first is Cross-task Preference Alignment (CPA), which transfers the preferences between tasks via optimal transport. The key idea of CPA is to use Gromov-Wasserstein distance to align the trajectories between tasks, and the solved optimal transport matrix serves as the correspondence between trajectories. The target task preferences are computed as the weighted sum of source task preference labels with the correspondence as weights. Moreover, to ensure robust learning from these transferred labels, we introduce Robust Reward Learning (RRL), which considers both reward mean and uncertainty by modeling rewards as Gaussian distributions. Empirical results on robotic manipulation tasks from Meta-World and Robomimic demonstrate that our method is capable of transferring preference labels across tasks accurately and then learns well-behaved policies. Notably, our approach significantly exceeds existing methods when there are few human preferences. The code and videos of our method are available at: https://sites.google.com/view/pearl-preference.
Authors: Ionut-Vlad Modoranu, Aleksei Kalinov, Eldar Kurtic, Elias Frantar, Dan Alistarh
Abstract: Leveraging second-order information about the loss at the scale of deep networks is one of the main lines of approach for improving the performance of current optimizers for deep learning. Yet, existing approaches for accurate full-matrix preconditioning, such as Full-Matrix Adagrad (GGT) or Matrix-Free Approximate Curvature (M-FAC) suffer from massive storage costs when applied even to small-scale models, as they must store a sliding window of gradients, whose memory requirements are multiplicative in the model dimension. In this paper, we address this issue via a novel and efficient error-feedback technique that can be applied to compress preconditioners by up to two orders of magnitude in practice, without loss of convergence. Specifically, our approach compresses the gradient information via sparsification or low-rank compression \emph{before} it is fed into the preconditioner, feeding the compression error back into future iterations. Experiments on deep neural networks show that this approach can compress full-matrix preconditioners to up to 99\% sparsity without accuracy loss, effectively removing the memory overhead of full-matrix preconditioners such as GGT and M-FAC. Our code is available at \url{https://github.com/IST-DASLab/EFCP}.
Authors: Xiner Li, Shurui Gui, Youzhi Luo, Shuiwang Ji
Abstract: Out-of-distribution (OOD) generalization deals with the prevalent learning scenario where test distribution shifts from training distribution. With rising application demands and inherent complexity, graph OOD problems call for specialized solutions. While data-centric methods exhibit performance enhancements on many generic machine learning tasks, there is a notable absence of data augmentation methods tailored for graph OOD generalization. In this work, we propose to achieve graph OOD generalization with the novel design of non-Euclidean-space linear extrapolation. The proposed augmentation strategy extrapolates both structure and feature spaces to generate OOD graph data. Our design tailors OOD samples for specific shifts without corrupting underlying causal mechanisms. Theoretical analysis and empirical results evidence the effectiveness of our method in solving target shifts, showing substantial and constant improvements across various graph OOD tasks.
Authors: Antonio Orvieto, Soham De, Caglar Gulcehre, Razvan Pascanu, Samuel L. Smith
Abstract: Deep neural networks based on linear RNNs interleaved with position-wise MLPs are gaining traction as competitive approaches for sequence modeling. Examples of such architectures include state-space models (SSMs) like S4, LRU, and Mamba: recently proposed models that achieve promising performance on text, genetics, and other data that require long-range reasoning. Despite experimental evidence highlighting these architectures' effectiveness and computational efficiency, their expressive power remains relatively unexplored, especially in connection to specific choices crucial in practice - e.g., carefully designed initialization distribution and potential use of complex numbers. In this paper, we show that combining MLPs with both real or complex linear diagonal recurrences leads to arbitrarily precise approximation of regular causal sequence-to-sequence maps. At the heart of our proof, we rely on a separation of concerns: the linear RNN provides a lossless encoding of the input sequence, and the MLP performs non-linear processing on this encoding. While we show that real diagonal linear recurrences are enough to achieve universality in this architecture, we prove that employing complex eigenvalues near unit disk - i.e., empirically the most successful strategy in S4 - greatly helps the RNN in storing information. We connect this finding with the vanishing gradient issue and provide experiments supporting our claims.
Authors: Hao Lin, Ke Wu, Jie Li, Jun Li, Wu-Jun Li
Abstract: Distributed learning is commonly used for training deep learning models, especially large models. In distributed learning, manual parallelism (MP) methods demand considerable human effort and have limited flexibility. Hence, automatic parallelism (AP) methods have recently been proposed for automating the parallel strategy optimization process. Existing AP methods suffer from sub-optimal solutions because they do not jointly optimize the two categories of parallel strategies (i.e., inter-layer parallelism and intra-layer parallelism). In this paper, we propose a novel AP method called UniAP, which unifies inter- and intra-layer automatic parallelism by mixed integer quadratic programming. To the best of our knowledge, UniAP is the first parallel method that can jointly optimize the two categories of parallel strategies to find an optimal solution. Experimental results show that UniAP outperforms state-of-the-art methods by up to 3.80$\times$ in throughput and reduces strategy optimization time by up to 107$\times$ across five Transformer-based models.
Authors: Nikita Balabin, Daria Voronkova, Ilya Trofimov, Evgeny Burnaev, Serguei Barannikov
Abstract: We propose TopDis (Topological Disentanglement), a method for learning disentangled representations via adding a multi-scale topological loss term. Disentanglement is a crucial property of data representations substantial for the explainability and robustness of deep learning models and a step towards high-level cognition. The state-of-the-art methods are based on VAE and encourage the joint distribution of latent variables to be factorized. We take a different perspective on disentanglement by analyzing topological properties of data manifolds. In particular, we optimize the topological similarity for data manifolds traversals. To the best of our knowledge, our paper is the first one to propose a differentiable topological loss for disentanglement learning. Our experiments have shown that the proposed TopDis loss improves disentanglement scores such as MIG, FactorVAE score, SAP score, and DCI disentanglement score with respect to state-of-the-art results while preserving the reconstruction quality. Our method works in an unsupervised manner, permitting us to apply it to problems without labeled factors of variation. The TopDis loss works even when factors of variation are correlated. Additionally, we show how to use the proposed topological loss to find disentangled directions in a trained GAN.
Authors: Sara Narteni, Alberto Carlevaro, Fabrizio Dabbene, Marco Muselli, Maurizio Mongelli
Abstract: Everyday life is increasingly influenced by artificial intelligence, and there is no question that machine learning algorithms must be designed to be reliable and trustworthy for everyone. Specifically, computer scientists consider an artificial intelligence system safe and trustworthy if it fulfills five pillars: explainability, robustness, transparency, fairness, and privacy. In addition to these five, we propose a sixth fundamental aspect: conformity, that is, the probabilistic assurance that the system will behave as the machine learner expects. In this paper, we present a methodology to link conformal prediction with explainable machine learning by defining a new score function for rule-based classifiers that leverages rules predictive ability, the geometrical position of points within rules boundaries and the overlaps among rules as well, thanks to the definition of a geometrical rule similarity term. Furthermore, we address the problem of defining regions in the feature space where conformal guarantees are satisfied, by exploiting the definition of conformal critical set and showing how this set can be used to achieve new rules with improved performance on the target class. The overall methodology is tested with promising results on several datasets of real-world interest, such as domain name server tunneling detection or cardiovascular disease prediction.
Authors: Nicolas Michel, Maorong Wang, Ling Xiao, Toshihiko Yamasaki
Abstract: Online Continual Learning (OCL) addresses the problem of training neural networks on a continuous data stream where multiple classification tasks emerge in sequence. In contrast to offline Continual Learning, data can be seen only once in OCL, which is a very severe constraint. In this context, replay-based strategies have achieved impressive results and most state-of-the-art approaches heavily depend on them. While Knowledge Distillation (KD) has been extensively used in offline Continual Learning, it remains under-exploited in OCL, despite its high potential. In this paper, we analyze the challenges in applying KD to OCL and give empirical justifications. We introduce a direct yet effective methodology for applying Momentum Knowledge Distillation (MKD) to many flagship OCL methods and demonstrate its capabilities to enhance existing approaches. In addition to improving existing state-of-the-art accuracy by more than $10\%$ points on ImageNet100, we shed light on MKD internal mechanics and impacts during training in OCL. We argue that similar to replay, MKD should be considered a central component of OCL. The code is available at \url{https://github.com/Nicolas1203/mkd_ocl}.
Authors: Arjun Subramonian, Levent Sagun, Yizhou Sun
Abstract: Graph neural network (GNN) link prediction is increasingly deployed in citation, collaboration, and online social networks to recommend academic literature, collaborators, and friends. While prior research has investigated the dyadic fairness of GNN link prediction, the within-group (e.g., queer women) fairness and "rich get richer" dynamics of link prediction remain underexplored. However, these aspects have significant consequences for degree and power imbalances in networks. In this paper, we shed light on how degree bias in networks affects Graph Convolutional Network (GCN) link prediction. In particular, we theoretically uncover that GCNs with a symmetric normalized graph filter have a within-group preferential attachment bias. We validate our theoretical analysis on real-world citation, collaboration, and online social networks. We further bridge GCN's preferential attachment bias with unfairness in link prediction and propose a new within-group fairness metric. This metric quantifies disparities in link prediction scores within social groups, towards combating the amplification of degree and power disparities. Finally, we propose a simple training-time strategy to alleviate within-group unfairness, and we show that it is effective on citation, social, and credit networks.
Authors: Alexander Mattick, Christopher Mutschler
Abstract: A big challenge in branch and bound lies in identifying the optimal node within the search tree from which to proceed. Current state-of-the-art selectors utilize either hand-crafted ensembles that automatically switch between naive sub-node selectors, or learned node selectors that rely on individual node data. We propose a novel simulation technique that uses reinforcement learning (RL) while considering the entire tree state, rather than just isolated nodes. To achieve this, we train a graph neural network that produces a probability distribution based on the path from the model's root to its "to-be-selected" leaves. Modelling node-selection as a probability distribution allows us to train the model using state-of-the-art RL techniques that capture both intrinsic node-quality and node-evaluation costs. Our method induces a high quality node selection policy on a set of varied and complex problem sets, despite only being trained on specially designed, synthetic travelling salesmen problem (TSP) instances. Using such a fixed pretrained policy shows significant improvements on several benchmarks in optimality gap reductions and per-node efficiency under strict time constraints.
Authors: Haoyu Ma, Jialong Wu, Ningya Feng, Chenjun Xiao, Dong Li, Jianye Hao, Jianmin Wang, Mingsheng Long
Abstract: Model-based reinforcement learning (MBRL) holds the promise of sample-efficient learning by utilizing a world model, which models how the environment works and typically encompasses components for two tasks: observation modeling and reward modeling. In this paper, through a dedicated empirical investigation, we gain a deeper understanding of the role each task plays in world models and uncover the overlooked potential of sample-efficient MBRL by mitigating the domination of either observation or reward modeling. Our key insight is that while prevalent approaches of explicit MBRL attempt to restore abundant details of the environment via observation models, it is difficult due to the environment's complexity and limited model capacity. On the other hand, reward models, while dominating implicit MBRL and adept at learning compact task-centric dynamics, are inadequate for sample-efficient learning without richer learning signals. Motivated by these insights and discoveries, we propose a simple yet effective approach, HarmonyDream, which automatically adjusts loss coefficients to maintain task harmonization, i.e. a dynamic equilibrium between the two tasks in world model learning. Our experiments show that the base MBRL method equipped with HarmonyDream gains 10%-69% absolute performance boosts on visual robotic tasks and sets a new state-of-the-art result on the Atari 100K benchmark. Code is available at https://github.com/thuml/HarmonyDream.
Authors: Tamara Cucumides, Daniel Daza, Pablo Barcel\'o, Michael Cochez, Floris Geerts, Juan L Reutter, Miguel Romero
Abstract: The challenge of answering graph queries over incomplete knowledge graphs is gaining significant attention in the machine learning community. Neuro-symbolic models have emerged as a promising approach, combining good performance with high interpretability. These models utilize trained architectures to execute atomic queries and integrate modules that mimic symbolic query operators. However, most neuro-symbolic query processors are constrained to tree-like graph pattern queries. These queries admit a bottom-up execution with constant values or anchors at the leaves and the target variable at the root. While expressive, tree-like queries fail to capture critical properties in knowledge graphs, such as the existence of multiple edges between entities or the presence of triangles. We introduce a framework for answering arbitrary graph pattern queries over incomplete knowledge graphs, encompassing both cyclic queries and tree-like queries with existentially quantified leaves. These classes of queries are vital for practical applications but are beyond the scope of most current neuro-symbolic models. Our approach employs an approximation scheme that facilitates acyclic traversals for cyclic patterns, thereby embedding additional symbolic bias into the query execution process. Our experimental evaluation demonstrates that our framework performs competitively on three datasets, effectively handling cyclic queries through our approximation strategy. Additionally, it maintains the performance of existing neuro-symbolic models on anchored tree-like queries and extends their capabilities to queries with existentially quantified variables.
Authors: Ying Sheng, Shiyi Cao, Dacheng Li, Coleman Hooper, Nicholas Lee, Shuo Yang, Christopher Chou, Banghua Zhu, Lianmin Zheng, Kurt Keutzer, Joseph E. Gonzalez, Ion Stoica
Abstract: The "pretrain-then-finetune" paradigm is commonly adopted in the deployment of large language models. Low-Rank Adaptation (LoRA), a parameter-efficient fine-tuning method, is often employed to adapt a base model to a multitude of tasks, resulting in a substantial collection of LoRA adapters derived from one base model. We observe that this paradigm presents significant opportunities for batched inference during serving. To capitalize on these opportunities, we present S-LoRA, a system designed for the scalable serving of many LoRA adapters. S-LoRA stores all adapters in the main memory and fetches the adapters used by the currently running queries to the GPU memory. To efficiently use the GPU memory and reduce fragmentation, S-LoRA proposes Unified Paging. Unified Paging uses a unified memory pool to manage dynamic adapter weights with different ranks and KV cache tensors with varying sequence lengths. Additionally, S-LoRA employs a novel tensor parallelism strategy and highly optimized custom CUDA kernels for heterogeneous batching of LoRA computation. Collectively, these features enable S-LoRA to serve thousands of LoRA adapters on a single GPU or across multiple GPUs with a small overhead. Compared to state-of-the-art libraries such as HuggingFace PEFT and vLLM (with naive support of LoRA serving), S-LoRA can improve the throughput by up to 4 times and increase the number of served adapters by several orders of magnitude. As a result, S-LoRA enables scalable serving of many task-specific fine-tuned models and offers the potential for large-scale customized fine-tuning services. The code is available at https://github.com/S-LoRA/S-LoRA
Authors: Shida Wang, Qianxiao Li
Abstract: In this paper, we investigate the long-term memory learning capabilities of state-space models (SSMs) from the perspective of parameterization. We prove that state-space models without any reparameterization exhibit a memory limitation similar to that of traditional RNNs: the target relationships that can be stably approximated by state-space models must have an exponential decaying memory. Our analysis identifies this "curse of memory" as a result of the recurrent weights converging to a stability boundary, suggesting that a reparameterization technique can be effective. To this end, we introduce a class of reparameterization techniques for SSMs that effectively lift its memory limitations. Besides improving approximation capabilities, we further illustrate that a principled choice of reparameterization scheme can also enhance optimization stability. We validate our findings using synthetic datasets, language models and image classifications.
Authors: Difan Jiao, Yilun Liu, Zhenwei Tang, Daniel Matter, J\"urgen Pfeffer, Ashton Anderson
Abstract: Among the many tasks that Large Language Models (LLMs) have revolutionized is text classification. Current text classification paradigms, however, rely solely on the output of the final layer in the LLM, with the rich information contained in internal neurons largely untapped. In this study, we present SPIN: a model-agnostic framework that sparsifies and integrates internal neurons of intermediate layers of LLMs for text classification. Specifically, SPIN sparsifies internal neurons by linear probing-based salient neuron selection layer by layer, avoiding noise from unrelated neurons and ensuring efficiency. The cross-layer salient neurons are then integrated to serve as multi-layered features for the classification head. Extensive experimental results show our proposed SPIN significantly improves text classification accuracy, efficiency, and interpretability.
Authors: Yuhui Ding, Antonio Orvieto, Bobby He, Thomas Hofmann
Abstract: Graph neural networks based on iterative one-hop message passing have been shown to struggle in harnessing the information from distant nodes effectively. Conversely, graph transformers allow each node to attend to all other nodes directly, but lack graph inductive bias and have to rely on ad-hoc positional encoding. In this paper, we propose a new architecture to reconcile these challenges. Our approach stems from the recent breakthroughs in long-range modeling provided by deep state-space models: for a given target node, our model aggregates other nodes by their shortest distances to the target and uses a linear RNN to encode the sequence of hop representations. The linear RNN is parameterized in a particular diagonal form for stable long-range signal propagation and is theoretically expressive enough to encode the neighborhood hierarchy. With no need for positional encoding, we empirically show that the performance of our model is comparable to or better than that of state-of-the-art graph transformers on various benchmarks, with a significantly reduced computational cost. Our code is open-source at https://github.com/skeletondyh/GRED.
Authors: Dan Friedman, Andrew Lampinen, Lucas Dixon, Danqi Chen, Asma Ghandeharioun
Abstract: A common method to study deep learning systems is to use simplified model representations--for example, using singular value decomposition to visualize the model's hidden states in a lower dimensional space. This approach assumes that the results of these simplifications are faithful to the original model. Here, we illustrate an important caveat to this assumption: even if the simplified representations can accurately approximate the full model on the training set, they may fail to accurately capture the model's behavior out of distribution. We illustrate this by training Transformer models on controlled datasets with systematic generalization splits, including the Dyck balanced-parenthesis languages and a code completion task. We simplify these models using tools like dimensionality reduction and clustering, and then explicitly test how these simplified proxies match the behavior of the original model. We find consistent generalization gaps: cases in which the simplified proxies are more faithful to the original model on the in-distribution evaluations and less faithful on various tests of systematic generalization. This includes cases where the original model generalizes systematically but the simplified proxies fail, and cases where the simplified proxies generalize better. Together, our results raise questions about the extent to which mechanistic interpretations derived using tools like SVD can reliably predict what a model will do in novel situations.
Authors: Songlin Yang, Bailin Wang, Yikang Shen, Rameswar Panda, Yoon Kim
Abstract: Transformers with linear attention allow for efficient parallel training but can simultaneously be formulated as an RNN with 2D (matrix-valued) hidden states, thus enjoying linear-time inference complexity. However, linear attention generally underperforms ordinary softmax attention. Moreover, current implementations of linear attention lack I/O-awareness and are thus slower than highly optimized implementations of softmax attention. This work describes a hardware-efficient algorithm for linear attention that trades off memory movement against parallelizability. The resulting implementation, dubbed FLASHLINEARATTENTION, is faster than FLASHATTENTION-2 (Dao, 2023) as a standalone layer even on short sequence lengths (e.g., 1K). We then generalize this algorithm to a more expressive variant of linear attention with data-dependent gates. When used as a replacement for the standard attention layer in Transformers, the resulting gated linear attention (GLA) Transformer is found to perform competitively against the LLaMA-architecture Transformer (Touvron et al., 2023) as well recent linear-time-inference baselines such as RetNet (Sun et al., 2023a) and Mamba (Gu & Dao, 2023) on moderate-scale language modeling experiments. GLA Transformer is especially effective at length generalization, enabling a model trained on 2K to generalize to sequences longer than 20K without significant perplexity degradations. For training speed, the GLA Transformer has higher throughput than a similarly-sized Mamba model.
Authors: Amin Yousefpour, Zahra Zanjani Foumani, Mehdi Shishehbor, Carlos Mora, Ramin Bostanabad
Abstract: In this paper we introduce GP+, an open-source library for kernel-based learning via Gaussian processes (GPs) which are powerful statistical models that are completely characterized by their parametric covariance and mean functions. GP+ is built on PyTorch and provides a user-friendly and object-oriented tool for probabilistic learning and inference. As we demonstrate with a host of examples, GP+ has a few unique advantages over other GP modeling libraries. We achieve these advantages primarily by integrating nonlinear manifold learning techniques with GPs' covariance and mean functions. As part of introducing GP+, in this paper we also make methodological contributions that (1) enable probabilistic data fusion and inverse parameter estimation, and (2) equip GPs with parsimonious parametric mean functions which span mixed feature spaces that have both categorical and quantitative variables. We demonstrate the impact of these contributions in the context of Bayesian optimization, multi-fidelity modeling, sensitivity analysis, and calibration of computer models.
Authors: Srishti Gautam, Ahcene Boubekki, Marina M. C. H\"ohne, Michael C. Kampffmeyer
Abstract: Explainable AI (XAI) has unfolded in two distinct research directions with, on the one hand, post-hoc methods that explain the predictions of a pre-trained black-box model and, on the other hand, self-explainable models (SEMs) which are trained directly to provide explanations alongside their predictions. While the latter is preferred in safety-critical scenarios, post-hoc approaches have received the majority of attention until now, owing to their simplicity and ability to explain base models without retraining. Current SEMs, instead, require complex architectures and heavily regularized loss functions, thus necessitating specific and costly training. To address this shortcoming and facilitate wider use of SEMs, we propose a simple yet efficient universal method called KMEx (K-Means Explainer), which can convert any existing pre-trained model into a prototypical SEM. The motivation behind KMEx is to enhance transparency in deep learning-based decision-making via class-prototype-based explanations that are diverse and trustworthy without retraining the base model. We compare models obtained from KMEx to state-of-the-art SEMs using an extensive qualitative evaluation to highlight the strengths and weaknesses of each model, further paving the way toward a more reliable and objective evaluation of SEMs (The code is available at https://github.com/SrishtiGautam/KMEx).
Authors: Ning Liu, Yiming Fan, Xianyi Zeng, Milan Kl\"ower, Lu Zhang, Yue Yu
Abstract: Neural operators (NOs) have emerged as effective tools for modeling complex physical systems in scientific machine learning. In NOs, a central characteristic is to learn the governing physical laws directly from data. In contrast to other machine learning applications, partial knowledge is often known a priori about the physical system at hand whereby quantities such as mass, energy and momentum are exactly conserved. Currently, NOs have to learn these conservation laws from data and can only approximately satisfy them due to finite training data and random noise. In this work, we introduce conservation law-encoded neural operators (clawNOs), a suite of NOs that endow inference with automatic satisfaction of such conservation laws. ClawNOs are built with a divergence-free prediction of the solution field, with which the continuity equation is automatically guaranteed. As a consequence, clawNOs are compliant with the most fundamental and ubiquitous conservation laws essential for correct physical consistency. As demonstrations, we consider a wide variety of scientific applications ranging from constitutive modeling of material deformation, incompressible fluid dynamics, to atmospheric simulation. ClawNOs significantly outperform the state-of-the-art NOs in learning efficacy, especially in small-data regimes.
Authors: Subham Sekhar Sahoo, Aaron Gokaslan, Chris De Sa, Volodymyr Kuleshov
Abstract: Diffusion models have gained traction as powerful algorithms for synthesizing high-quality images. Central to these algorithms is the diffusion process, a set of equations which maps data to noise in a way that can significantly affect performance. In this paper, we explore whether the diffusion process can be learned from data. Our work is grounded in Bayesian inference and seeks to improve log-likelihood estimation by casting the learned diffusion process as an approximate variational posterior that yields a tighter lower bound (ELBO) on the likelihood. A widely held assumption is that the ELBO is invariant to the noise process: our work dispels this assumption and proposes multivariate learned adaptive noise (MULAN), a learned diffusion process that applies noise at different rates across an image. Specifically, our method relies on a multivariate noise schedule that is a function of the data to ensure that the ELBO is no longer invariant to the choice of the noise schedule as in previous works. Empirically, MULAN sets a new state-of-the-art in density estimation on CIFAR-10 and ImageNet and reduces the number of training steps by 50%. Code is available at https://github.com/s-sahoo/MuLAN
Authors: Laura Sisson, Aryan Amit Barsainyan, Mrityunjay Sharma, Ritesh Kumar
Abstract: The application of deep learning techniques on aroma-chemicals has resulted in models more accurate than human experts at predicting olfactory qualities. However, public research in this domain has been limited to predicting the qualities of single molecules, whereas in industry applications, perfumers and food scientists are often concerned with blends of many molecules. In this paper, we apply both existing and novel approaches to a dataset we gathered consisting of labeled pairs of molecules. We present graph neural network models capable of accurately predicting the odor qualities arising from blends of aroma-chemicals, with an analysis of how variations in architecture can lead to significant differences in predictive power.
Authors: Victor Dheur, Tanguy Bosser, Rafael Izbicki, Souhaib Ben Taieb
Abstract: Sequences of labeled events observed at irregular intervals in continuous time are ubiquitous across various fields. Temporal Point Processes (TPPs) provide a mathematical framework for modeling these sequences, enabling inferences such as predicting the arrival time of future events and their associated label, called mark. However, due to model misspecification or lack of training data, these probabilistic models may provide a poor approximation of the true, unknown underlying process, with prediction regions extracted from them being unreliable estimates of the underlying uncertainty. This paper develops more reliable methods for uncertainty quantification in neural TPP models via the framework of conformal prediction. A primary objective is to generate a distribution-free joint prediction region for an event's arrival time and mark, with a finite-sample marginal coverage guarantee. A key challenge is to handle both a strictly positive, continuous response and a categorical response, without distributional assumptions. We first consider a simple but conservative approach that combines individual prediction regions for the event's arrival time and mark. Then, we introduce a more effective method based on bivariate highest density regions derived from the joint predictive density of arrival times and marks. By leveraging the dependencies between these two variables, this method excludes unlikely combinations of the two, resulting in sharper prediction regions while still attaining the pre-specified coverage level. We also explore the generation of individual univariate prediction regions for events' arrival times and marks through conformal regression and classification techniques. Moreover, we evaluate the stronger notion of conditional coverage. Finally, through extensive experimentation on both simulated and real-world datasets, we assess the validity and efficiency of these methods.
Authors: Kaiyi Zhang, Ang Lv, Yuhan Chen, Hansen Ha, Tao Xu, Rui Yan
Abstract: In this paper, by treating in-context learning (ICL) as a meta-optimization process, we explain why LLMs are sensitive to the order of ICL examples. This understanding leads us to the development of Batch-ICL, an effective, efficient, and order-agnostic inference algorithm for ICL. Differing from the standard N-shot learning approach, Batch-ICL employs $N$ separate 1-shot forward computations and aggregates the resulting meta-gradients. These aggregated meta-gradients are then applied to the forward computation of a zero-shot query to generate the final prediction. This batch processing approach renders the LLM agnostic to the order of ICL examples. Through extensive experiments and analysis, we demonstrate that Batch-ICL consistently outperforms most permutations of ICL examples. In some cases, it even exceeds the performance of the best order for standard ICL, all while reducing the computational resources required. Furthermore, we develop a novel variant of Batch-ICL featuring multiple "epochs" of meta-optimization. This variant implicitly explores permutations of ICL examples, further enhancing ICL performance.
Authors: M\'elanie Ducoffe, Guillaume Pov\'eda, Audrey Galametz, Ryma Boumazouza, Marion-C\'ecile Martin, Julien Baris, Derk Daverschot, Eugene O'Higgins
Abstract: Surrogate Neural Networks are nowadays routinely used in industry as substitutes for computationally demanding engineering simulations (e.g., in structural analysis). They allow to generate faster predictions and thus analyses in industrial applications e.g., during a product design, testing or monitoring phases. Due to their performance and time-efficiency, these surrogate models are now being developed for use in safety-critical applications. Neural network verification and in particular the assessment of their robustness (e.g., to perturbations) is the next critical step to allow their inclusion in real-life applications and certification. We assess the applicability and scalability of empirical and formal methods in the context of aircraft predictive maintenance for surrogate neural networks designed to predict the stress sustained by an aircraft part from external loads. The case study covers a high-dimensional input and output space and the verification process thus accommodates multi-objective constraints. We explore the complementarity of verification methods in assessing the local stability property of such surrogate models to input noise. We showcase the effectiveness of sequentially combining methods in one verification 'pipeline' and demonstrating the subsequent gain in runtime required to assess the targeted property.
Authors: Philip Amortila, Dylan J. Foster, Nan Jiang, Ayush Sekhari, Tengyang Xie
Abstract: The theories of offline and online reinforcement learning, despite having evolved in parallel, have begun to show signs of the possibility for a unification, with algorithms and analysis techniques for one setting often having natural counterparts in the other. However, the notion of density ratio modeling, an emerging paradigm in offline RL, has been largely absent from online RL, perhaps for good reason: the very existence and boundedness of density ratios relies on access to an exploratory dataset with good coverage, but the core challenge in online RL is to collect such a dataset without having one to start. In this work we show -- perhaps surprisingly -- that density ratio-based algorithms have online counterparts. Assuming only the existence of an exploratory distribution with good coverage, a structural condition known as coverability (Xie et al., 2023), we give a new algorithm (GLOW) that uses density ratio realizability and value function realizability to perform sample-efficient online exploration. GLOW addresses unbounded density ratios via careful use of truncation, and combines this with optimism to guide exploration. GLOW is computationally inefficient; we complement it with a more efficient counterpart, HyGLOW, for the Hybrid RL setting (Song et al., 2022) wherein online RL is augmented with additional offline data. HyGLOW is derived as a special case of a more general meta-algorithm that provides a provable black-box reduction from hybrid RL to offline RL, which may be of independent interest.
Authors: Qiong Wu, Wenhua Wang, Pingyi Fan, Qiang Fan, Huiling Zhu, Khaled B. Letaief
Abstract: Edge caching is a promising solution for next-generation networks by empowering caching units in small-cell base stations (SBSs), which allows user equipments (UEs) to fetch users' requested contents that have been pre-cached in SBSs. It is crucial for SBSs to predict accurate popular contents through learning while protecting users' personal information. Traditional federated learning (FL) can protect users' privacy but the data discrepancies among UEs can lead to a degradation in model quality. Therefore, it is necessary to train personalized local models for each UE to predict popular contents accurately. In addition, the cached contents can be shared among adjacent SBSs in next-generation networks, thus caching predicted popular contents in different SBSs may affect the cost to fetch contents. Hence, it is critical to determine where the popular contents are cached cooperatively. To address these issues, we propose a cooperative edge caching scheme based on elastic federated and multi-agent deep reinforcement learning (CEFMR) to optimize the cost in the network. We first propose an elastic FL algorithm to train the personalized model for each UE, where adversarial autoencoder (AAE) model is adopted for training to improve the prediction accuracy, then {a popular} content prediction algorithm is proposed to predict the popular contents for each SBS based on the trained AAE model. Finally, we propose a multi-agent deep reinforcement learning (MADRL) based algorithm to decide where the predicted popular contents are collaboratively cached among SBSs. Our experimental results demonstrate the superiority of our proposed scheme to existing baseline caching schemes.
Authors: Vikranth Dwaracherla, Seyed Mohammad Asghari, Botao Hao, Benjamin Van Roy
Abstract: We present evidence of substantial benefit from efficient exploration in gathering human feedback to improve large language models. In our experiments, an agent sequentially generates queries while fitting a reward model to the feedback received. Our best-performing agent generates queries using double Thompson sampling, with uncertainty represented by an epistemic neural network. Our results demonstrate that efficient exploration enables high levels of performance with far fewer queries. Further, both uncertainty estimation and the choice of exploration scheme play critical roles.
Authors: Tobias Fuchs, Florian Kalinke, Klemens B\"ohm
Abstract: In real-world applications, one often encounters ambiguously labeled data, where different annotators assign conflicting class labels. Partial-label learning allows training classifiers in this weakly supervised setting, where state-of-the-art methods already show good predictive performance. However, even the best algorithms give incorrect predictions, which can have severe consequences when they impact actions or decisions. We propose a novel risk-consistent partial-label learning algorithm with a reject option, that is, the algorithm can reject unsure predictions. Extensive experiments on artificial and real-world datasets show that our method provides the best trade-off between the number and accuracy of non-rejected predictions when compared to our competitors, which use confidence thresholds for rejecting unsure predictions instead. When evaluated without the reject option, our nearest neighbor-based approach also achieves competitive prediction performance.
Authors: Jack Foster, Kyle Fogarty, Stefan Schoepf, Cengiz \"Oztireli, Alexandra Brintrup
Abstract: To comply with AI and data regulations, the need to forget private or copyrighted information from trained machine learning models is increasingly important. The key challenge in unlearning is forgetting the necessary data in a timely manner, while preserving model performance. In this work, we address the zero-shot unlearning scenario, whereby an unlearning algorithm must be able to remove data given only a trained model and the data to be forgotten. We explore unlearning from an information theoretic perspective, connecting the influence of a sample to the information gain a model receives by observing it. From this, we derive a simple but principled zero-shot unlearning method based on the geometry of the model. Our approach takes the form of minimising the gradient of a learned function with respect to a small neighbourhood around a target forget point. This induces a smoothing effect, causing forgetting by moving the boundary of the classifier. We explore the intuition behind why this approach can jointly unlearn forget samples while preserving general model performance through a series of low-dimensional experiments. We perform extensive empirical evaluation of our method over a range of contemporary benchmarks, verifying that our method is competitive with state-of-the-art performance under the strict constraints of zero-shot unlearning.
Authors: Hao Chen, Jindong Wang, Lei Feng, Xiang Li, Yidong Wang, Xing Xie, Masashi Sugiyama, Rita Singh, Bhiksha Raj
Abstract: Weakly supervised learning generally faces challenges in applicability to various scenarios with diverse weak supervision and in scalability due to the complexity of existing algorithms, thereby hindering the practical deployment. This paper introduces a general framework for learning from weak supervision (GLWS) with a novel algorithm. Central to GLWS is an Expectation-Maximization (EM) formulation, adeptly accommodating various weak supervision sources, including instance partial labels, aggregate statistics, pairwise observations, and unlabeled data. We further present an advanced algorithm that significantly simplifies the EM computational demands using a Non-deterministic Finite Automaton (NFA) along with a forward-backward algorithm, which effectively reduces time complexity from quadratic or factorial often required in existing solutions to linear scale. The problem of learning from arbitrary weak supervision is therefore converted to the NFA modeling of them. GLWS not only enhances the scalability of machine learning models but also demonstrates superior performance and versatility across 11 weak supervision scenarios. We hope our work paves the way for further advancements and practical deployment in this field.
Authors: Christopher Morris, Fabrizio Frasca, Nadav Dym, Haggai Maron, \.Ismail \.Ilkan Ceylan, Ron Levie, Derek Lim, Michael Bronstein, Martin Grohe, Stefanie Jegelka
Abstract: Machine learning on graphs, especially using graph neural networks (GNNs), has seen a surge in interest due to the wide availability of graph data across a broad spectrum of disciplines, from life to social and engineering sciences. Despite their practical success, our theoretical understanding of the properties of GNNs remains highly incomplete. Recent theoretical advancements primarily focus on elucidating the coarse-grained expressive power of GNNs, predominantly employing combinatorial techniques. However, these studies do not perfectly align with practice, particularly in understanding the generalization behavior of GNNs when trained with stochastic first-order optimization techniques. In this position paper, we argue that the graph machine learning community needs to shift its attention to developing a balanced theory of graph machine learning, focusing on a more thorough understanding of the interplay of expressive power, generalization, and optimization.
Authors: Fangzhao Zhang, Mert Pilanci
Abstract: Low-Rank Adaptation (LoRA) emerges as a popular parameter-efficient fine-tuning (PEFT) method, which proposes to freeze pretrained model weights and update an additive low-rank trainable matrix. In this work, we study the enhancement of LoRA training by introducing an $r \times r$ preconditioner in each gradient step where $r$ is the LoRA rank. We theoretically verify that the proposed preconditioner stabilizes feature learning with LoRA under infinite-width NN setting. Empirically, the implementation of this new preconditioner requires a small change to existing optimizer code and creates virtually minuscule storage and runtime overhead. Our experimental results with both large language models and text-to-image diffusion models show that with this new preconditioner, the convergence and reliability of SGD and AdamW can be significantly enhanced. Moreover, the training process becomes much more robust to hyperparameter choices such as learning rate. The new preconditioner can be derived from a novel Riemannian metric in low-rank matrix field. Code can be accessed at https://github.com/pilancilab/Riemannian_Preconditioned_LoRA.
URLs: https://github.com/pilancilab/Riemannian_Preconditioned_LoRA.
Authors: Qilong Ma, Haixu Wu, Lanxiang Xing, Shangchen Miao, Mingsheng Long
Abstract: Accurately predicting the future fluid is vital to extensive areas such as meteorology, oceanology, and aerodynamics. However, since the fluid is usually observed from the Eulerian perspective, its moving and intricate dynamics are seriously obscured and confounded in static grids, bringing thorny challenges to the prediction. This paper introduces a new Lagrangian-Eulerian combined paradigm to tackle the tanglesome fluid dynamics. Instead of solely predicting the future based on Eulerian observations, we propose DeepLag to discover hidden Lagrangian dynamics within the fluid by tracking the movements of adaptively sampled key particles. Further, DeepLag presents a new paradigm for fluid prediction, where the Lagrangian movement of the tracked particles is inferred from Eulerian observations, and their accumulated Lagrangian dynamics information is incorporated into global Eulerian evolving features to guide future prediction respectively. Tracking key particles not only provides a transparent and interpretable clue for fluid dynamics but also makes our model free from modeling complex correlations among massive grids for better efficiency. Experimentally, DeepLag excels in three challenging fluid prediction tasks covering 2D and 3D, simulated and real-world fluids.
Authors: Samuel Garcin, James Doran, Shangmin Guo, Christopher G. Lucas, Stefano V. Albrecht
Abstract: Autonomous agents trained using deep reinforcement learning (RL) often lack the ability to successfully generalise to new environments, even when these environments share characteristics with the ones they have encountered during training. In this work, we investigate how the sampling of individual environment instances, or levels, affects the zero-shot generalisation (ZSG) ability of RL agents. We discover that, for deep actor-critic architectures sharing their base layers, prioritising levels according to their value loss minimises the mutual information between the agent's internal representation and the set of training levels in the generated training data. This provides a novel theoretical justification for the regularisation achieved by certain adaptive sampling strategies. We then turn our attention to unsupervised environment design (UED) methods, which assume control over level generation. We find that existing UED methods can significantly shift the training distribution, which translates to low ZSG performance. To prevent both overfitting and distributional shift, we introduce data-regularised environment design (DRED). DRED generates levels using a generative model trained to approximate the ground truth distribution of an initial set of level parameters. Through its grounding, DRED achieves significant improvements in ZSG over adaptive level sampling strategies and UED methods. Our code and experimental data are available at https://github.com/uoe-agents/dred.
Authors: Artem Vysogorets, Anna Dawid, Julia Kempe
Abstract: The second-order properties of the training loss have a massive impact on the optimization dynamics of deep learning models. Fort & Scherlis (2019) discovered that a large excess of positive curvature and local convexity of the loss Hessian is associated with highly trainable initial points located in a region coined the "Goldilocks zone". Only a handful of subsequent studies touched upon this relationship, so it remains largely unexplained. In this paper, we present a rigorous and comprehensive analysis of the Goldilocks zone for homogeneous neural networks. In particular, we derive the fundamental condition resulting in excess of positive curvature of the loss, explaining and refining its conventionally accepted connection to the initialization norm. Further, we relate the excess of positive curvature to model confidence, low initial loss, and a previously unknown type of vanishing cross-entropy loss gradient. To understand the importance of excessive positive curvature for trainability of deep networks, we optimize fully-connected and convolutional architectures outside the Goldilocks zone and analyze the emergent behaviors. We find that strong model performance is not perfectly aligned with the Goldilocks zone, calling for further research into this relationship.
Authors: Kelsey Lieberman, Shuai Yuan, Swarna Kamlam Ravindran, Carlo Tomasi
Abstract: Although binary classification is a well-studied problem in computer vision, training reliable classifiers under severe class imbalance remains a challenging problem. Recent work has proposed techniques that mitigate the effects of training under imbalance by modifying the loss functions or optimization methods. While this work has led to significant improvements in the overall accuracy in the multi-class case, we observe that slight changes in hyperparameter values of these methods can result in highly variable performance in terms of Receiver Operating Characteristic (ROC) curves on binary problems with severe imbalance. To reduce the sensitivity to hyperparameter choices and train more general models, we propose training over a family of loss functions, instead of a single loss function. We develop a method for applying Loss Conditional Training (LCT) to an imbalanced classification problem. Extensive experiment results, on both CIFAR and Kaggle competition datasets, show that our method improves model performance and is more robust to hyperparameter choices. Code is available at https://github.com/klieberman/roc_lct.
Authors: Felix Draxler, Stefan Wahl, Christoph Schn\"orr, Ullrich K\"othe
Abstract: We present a novel theoretical framework for understanding the expressive power of normalizing flows. Despite their prevalence in scientific applications, a comprehensive understanding of flows remains elusive due to their restricted architectures. Existing theorems fall short as they require the use of arbitrarily ill-conditioned neural networks, limiting practical applicability. We propose a distributional universality theorem for well-conditioned coupling-based normalizing flows such as RealNVP. In addition, we show that volume-preserving normalizing flows are not universal, what distribution they learn instead, and how to fix their expressivity. Our results support the general wisdom that affine and related couplings are expressive and in general outperform volume-preserving flows, bridging a gap between empirical results and theoretical understanding.
Authors: Ziyi Zhang, Sen Zhang, Yibing Zhan, Yong Luo, Yonggang Wen, Dacheng Tao
Abstract: Bridging the gap between diffusion models and human preferences is crucial for their integration into practical generative workflows. While optimizing downstream reward models has emerged as a promising alignment strategy, concerns arise regarding the risk of excessive optimization with learned reward models, which potentially compromises ground-truth performance. In this work, we confront the reward overoptimization problem in diffusion model alignment through the lenses of both inductive and primacy biases. We first identify a mismatch between current methods and the temporal inductive bias inherent in the multi-step denoising process of diffusion models, as a potential source of reward overoptimization. Then, we surprisingly discover that dormant neurons in our critic model act as a regularization against reward overoptimization while active neurons reflect primacy bias. Motivated by these observations, we propose Temporal Diffusion Policy Optimization with critic active neuron Reset (TDPO-R), a policy gradient algorithm that exploits the temporal inductive bias of diffusion models and mitigates the primacy bias stemming from active neurons. Empirical results demonstrate the superior efficacy of our methods in mitigating reward overoptimization. Code is avaliable at https://github.com/ZiyiZhang27/tdpo.
Authors: Juntao Ren, Gokul Swamy, Zhiwei Steven Wu, J. Andrew Bagnell, Sanjiban Choudhury
Abstract: The inverse reinforcement learning approach to imitation learning is a double-edged sword. On the one hand, it can enable learning from a smaller number of expert demonstrations with more robustness to error compounding than behavioral cloning approaches. On the other hand, it requires that the learner repeatedly solve a computationally expensive reinforcement learning (RL) problem. Often, much of this computation is wasted searching over policies very dissimilar to the expert's. In this work, we propose using hybrid RL -- training on a mixture of online and expert data -- to curtail unnecessary exploration. Intuitively, the expert data focuses the learner on good states during training, which reduces the amount of exploration required to compute a strong policy. Notably, such an approach doesn't need the ability to reset the learner to arbitrary states in the environment, a requirement of prior work in efficient inverse RL. More formally, we derive a reduction from inverse RL to expert-competitive RL (rather than globally optimal RL) that allows us to dramatically reduce interaction during the inner policy search loop while maintaining the benefits of the IRL approach. This allows us to derive both model-free and model-based hybrid inverse RL algorithms with strong policy performance guarantees. Empirically, we find that our approaches are significantly more sample efficient than standard inverse RL and several other baselines on a suite of continuous control tasks.
Authors: Shashwat Singh, Shauli Ravfogel, Jonathan Herzig, Roee Aharoni, Ryan Cotterell, Ponnurangam Kumaraguru
Abstract: Language models often exhibit undesirable behavior, e.g., generating toxic or gender-biased text. In the case of neural language models, an encoding of the undesirable behavior is often present in the model's representations. Thus, one natural (and common) approach to prevent the model from exhibiting undesirable behavior is to steer the model's representations in a manner that reduces the probability of it generating undesirable text. This paper investigates the formal and empirical properties of steering functions, i.e., transformation of the neural language model's representations that alter its behavior. First, we derive two optimal, in the least-squares sense, affine steering functions under different constraints. Our theory provides justification for existing approaches and offers a novel, improved steering approach. Second, we offer a series of experiments that demonstrate the empirical effectiveness of the methods in mitigating bias and reducing toxic generation.
Authors: Rui Yang, Xiaoman Pan, Feng Luo, Shuang Qiu, Han Zhong, Dong Yu, Jianshu Chen
Abstract: We consider the problem of multi-objective alignment of foundation models with human preferences, which is a critical step towards helpful and harmless AI systems. However, it is generally costly and unstable to fine-tune large foundation models using reinforcement learning (RL), and the multi-dimensionality, heterogeneity, and conflicting nature of human preferences further complicate the alignment process. In this paper, we introduce Rewards-in-Context (RiC), which conditions the response of a foundation model on multiple rewards in its prompt context and applies supervised fine-tuning for alignment. The salient features of RiC are simplicity and adaptivity, as it only requires supervised fine-tuning of a single foundation model and supports dynamic adjustment for user preferences during inference time. Inspired by the analytical solution of an abstracted convex optimization problem, our dynamic inference-time adjustment method approaches the Pareto-optimal solution for multiple objectives. Empirical evidence demonstrates the efficacy of our method in aligning both Large Language Models (LLMs) and diffusion models to accommodate diverse rewards with only around 10% GPU hours compared with multi-objective RL baseline.
Authors: Nirjhar Das, Souradip Chakraborty, Aldo Pacchiano, Sayak Ray Chowdhury
Abstract: Reinforcement Learning from Human Feedback (RLHF) is pivotal in aligning Large Language Models (LLMs) with human preferences. Although aligned generative models have shown remarkable abilities in various tasks, their reliance on high-quality human preference data creates a costly bottleneck in the practical application of RLHF. One primary reason is that current methods rely on uniformly picking prompt-generation pairs from a dataset of prompt-generations, to collect human feedback, resulting in sub-optimal alignment under a constrained budget, which highlights the criticality of adaptive strategies in efficient alignment. Recent works [Mehta et al., 2023, Muldrew et al., 2024] have tried to address this problem by designing various heuristics based on generation uncertainty. However, either the assumptions in [Mehta et al., 2023] are restrictive, or [Muldrew et al., 2024] do not provide any rigorous theoretical guarantee. To address these, we reformulate RLHF within contextual preference bandit framework, treating prompts as contexts, and develop an active-learning algorithm, $\textit{Active Preference Optimization}$ ($\texttt{APO}$), which enhances model alignment by querying preference data from the most important samples, achieving superior performance for small sample budget. We analyze the theoretical performance guarantees of $\texttt{APO}$ under the BTL preference model showing that the suboptimality gap of the policy learned via $\texttt{APO}$ scales as $O(1/\sqrt{T})$ for a budget of $T$. We also show that collecting preference data by choosing prompts randomly leads to a policy that suffers a constant sub-optimality. We perform detailed experimental evaluations on practical preference datasets to validate $\texttt{APO}$'s efficacy over the existing methods, establishing it as a sample-efficient and practical solution of alignment in a cost-effective and scalable manner.
Authors: Yaroslav Aksenov, Nikita Balagansky, Sofia Maria Lo Cicero Vaina, Boris Shaposhnikov, Alexey Gorbatovski, Daniil Gavrilov
Abstract: Advancing the frontier of subquadratic architectures for Language Models (LMs) is crucial in the rapidly evolving field of natural language processing. Current innovations, including State Space Models, were initially celebrated for surpassing Transformer performance on language modeling tasks. However, these models have revealed deficiencies in essential In-Context Learning capabilities - a domain where the Transformer traditionally shines. The Based model emerged as a hybrid solution, blending a Linear Transformer with a kernel inspired by the Taylor expansion of exponential functions, augmented by convolutional networks. Mirroring the Transformer's in-context adeptness, it became a strong contender in the field. In our work, we present a singular, elegant alteration to the Based kernel that amplifies its In-Context Learning abilities evaluated with the Multi-Query Associative Recall task and overall language modeling process, as demonstrated on the Pile dataset.
Authors: Amit Dhurandhar, Swagatam Haldar, Dennis Wei, Karthikeyan Natesan Ramamurthy
Abstract: Given the black box nature of machine learning models, a plethora of explainability methods have been developed to decipher the factors behind individual decisions. In this paper, we introduce a novel problem of black box (probabilistic) explanation certification. We ask the question: Given a black box model with only query access, an explanation for an example and a quality metric (viz. fidelity, stability), can we find the largest hypercube (i.e., $\ell_{\infty}$ ball) centered at the example such that when the explanation is applied to all examples within the hypercube, (with high probability) a quality criterion is met (viz. fidelity greater than some value)? Being able to efficiently find such a \emph{trust region} has multiple benefits: i) insight into model behavior in a \emph{region}, with a \emph{guarantee}; ii) ascertained \emph{stability} of the explanation; iii) \emph{explanation reuse}, which can save time, energy and money by not having to find explanations for every example; and iv) a possible \emph{meta-metric} to compare explanation methods. Our contributions include formalizing this problem, proposing solutions, providing theoretical guarantees for these solutions that are computable, and experimentally showing their efficacy on synthetic and real data.
Authors: Zhixun Chen, Yali Du, David Mguni
Abstract: Many leading language models (LMs) use high-intensity computational resources both during training and execution. This poses the challenge of lowering resource costs for deployment and faster execution of decision-making tasks among others. We introduce a novel plug-and-play LM framework named Language Optimising Network Distribution (LONDI) framework. LONDI learns to selectively employ large LMs only where complex decision-making and reasoning are required while using low-resource LMs (i.e. LMs require less GPU usage, but may not be able to solve the problem alone) everywhere else. LONDI consists of a system of two (off-)policy networks, an LM, a large LM (LLM), and a reinforcement learning module that uses switching controls to quickly learn which system states to call the LLM. We then introduce a variant of LONDI that maintains budget constraints on LLM calls and hence its resource usage. Theoretically, we prove LONDI learns the subset of system states to activate the LLM required to solve the task. We then prove that LONDI converges to optimal solutions while also preserving budgetary constraints on LLM calls almost surely enabling it to solve various tasks while significantly lowering computational costs. We test LONDI's performance in a range of tasks in ScienceWorld and BabyAI-Text and demonstrate that LONDI can solve tasks only solvable by resource-intensive LLMs while reducing GPU usage by up to 30%.
Authors: Christian Schlarmann, Naman Deep Singh, Francesco Croce, Matthias Hein
Abstract: Multi-modal foundation models like OpenFlamingo, LLaVA, and GPT-4 are increasingly used for various real-world tasks. Prior work has shown that these models are highly vulnerable to adversarial attacks on the vision modality. These attacks can be leveraged to spread fake information or defraud users, and thus pose a significant risk, which makes the robustness of large multi-modal foundation models a pressing problem. The CLIP model, or one of its variants, is used as a frozen vision encoder in many large vision-language models (LVLMs), e.g. LLaVA and OpenFlamingo. We propose an unsupervised adversarial fine-tuning scheme to obtain a robust CLIP vision encoder, which yields robustness on all vision down-stream tasks (LVLMs, zero-shot classification) that rely on CLIP. In particular, we show that stealth-attacks on users of LVLMs by a malicious third party providing manipulated images are no longer possible once one replaces the original CLIP model with our robust one. No retraining or fine-tuning of the down-stream LVLMs is required. The code and robust models are available at https://github.com/chs20/RobustVLM
Authors: Siqi Miao, Zhiyuan Lu, Mia Liu, Javier Duarte, Pan Li
Abstract: This study introduces a novel transformer model optimized for large-scale point cloud processing in scientific domains such as high-energy physics (HEP) and astrophysics. Addressing the limitations of graph neural networks and standard transformers, our model integrates local inductive bias and achieves near-linear complexity with hardware-friendly regular operations. One contribution of this work is the quantitative analysis of the error-complexity tradeoff of various sparsification techniques for building efficient transformers. Our findings highlight the superiority of using locality-sensitive hashing (LSH), especially OR & AND-construction LSH, in kernel approximation for large-scale point cloud data with local inductive bias. Based on this finding, we propose LSH-based Efficient Point Transformer (HEPT), which combines E$^2$LSH with OR & AND constructions and is built upon regular computations. HEPT demonstrates remarkable performance on two critical yet time-consuming HEP tasks, significantly outperforming existing GNNs and transformers in accuracy and computational speed, marking a significant advancement in geometric deep learning and large-scale scientific data processing. Our code is available at https://github.com/Graph-COM/HEPT.
Authors: Saul Santos, Vlad Niculae, Daniel McNamee, Andre F. T. Martins
Abstract: Modern Hopfield networks have enjoyed recent interest due to their connection to attention in transformers. Our paper provides a unified framework for sparse Hopfield networks by establishing a link with Fenchel-Young losses. The result is a new family of Hopfield-Fenchel-Young energies whose update rules are end-to-end differentiable sparse transformations. We reveal a connection between loss margins, sparsity, and exact memory retrieval. We further extend this framework to structured Hopfield networks via the SparseMAP transformation, which can retrieve pattern associations instead of a single pattern. Experiments on multiple instance learning and text rationalization demonstrate the usefulness of our approach.
Authors: Mitchell Black, Zhengchao Wan, Gal Mishne, Amir Nayyeri, Yusu Wang
Abstract: The distinguishing power of graph transformers is closely tied to the choice of positional encoding: features used to augment the base transformer with information about the graph. There are two primary types of positional encoding: absolute positional encodings (APEs) and relative positional encodings (RPEs). APEs assign features to each node and are given as input to the transformer. RPEs instead assign a feature to each pair of nodes, e.g., graph distance, and are used to augment the attention block. A priori, it is unclear which method is better for maximizing the power of the resulting graph transformer. In this paper, we aim to understand the relationship between these different types of positional encodings. Interestingly, we show that graph transformers using APEs and RPEs are equivalent in terms of distinguishing power. In particular, we demonstrate how to interchange APEs and RPEs while maintaining their distinguishing power in terms of graph transformers. Based on our theoretical results, we provide a study on several APEs and RPEs (including the resistance distance and the recently introduced stable and expressive positional encoding (SPE)) and compare their distinguishing power in terms of transformers. We believe our work will help navigate the huge number of choices of positional encoding and will provide guidance on the future design of positional encodings for graph transformers.
Authors: Yihao Zhang, Hangzhou He, Jingyu Zhu, Huanran Chen, Yifei Wang, Zeming Wei
Abstract: Adversarial Training (AT), which adversarially perturb the input samples during training, has been acknowledged as one of the most effective defenses against adversarial attacks, yet suffers from inevitably decreased clean accuracy. Instead of perturbing the samples, Sharpness-Aware Minimization (SAM) perturbs the model weights during training to find a more flat loss landscape and improve generalization. However, as SAM is designed for better clean accuracy, its effectiveness in enhancing adversarial robustness remains unexplored. In this work, considering the duality between SAM and AT, we investigate the adversarial robustness derived from SAM. Intriguingly, we find that using SAM alone can improve adversarial robustness. To understand this unexpected property of SAM, we first provide empirical and theoretical insights into how SAM can implicitly learn more robust features, and conduct comprehensive experiments to show that SAM can improve adversarial robustness notably without sacrificing any clean accuracy, shedding light on the potential of SAM to be a substitute for AT when accuracy comes at a higher priority. Code is available at https://github.com/weizeming/SAM_AT.
Authors: Hongkang Li, Meng Wang, Songtao Lu, Xiaodong Cui, Pin-Yu Chen
Abstract: Transformer-based large language models have displayed impressive in-context learning capabilities, where a pre-trained model can handle new tasks without fine-tuning by simply augmenting the query with some input-output examples from that task. Despite the empirical success, the mechanics of how to train a Transformer to achieve ICL and the corresponding ICL capacity is mostly elusive due to the technical challenges of analyzing the nonconvex training problems resulting from the nonlinear self-attention and nonlinear activation in Transformers. To the best of our knowledge, this paper provides the first theoretical analysis of the training dynamics of Transformers with nonlinear self-attention and nonlinear MLP, together with the ICL generalization capability of the resulting model. Focusing on a group of binary classification tasks, we train Transformers using data from a subset of these tasks and quantify the impact of various factors on the ICL generalization performance on the remaining unseen tasks with and without data distribution shifts. We also analyze how different components in the learned Transformers contribute to the ICL performance. Furthermore, we provide the first theoretical analysis of how model pruning affects ICL performance and prove that proper magnitude-based pruning can have a minimal impact on ICL while reducing inference costs. These theoretical findings are justified through numerical experiments.
Authors: Cong Zhang, Zhiguang Cao, Yaoxin Wu, Wen Song, Jing Sun
Abstract: Existing learning-based methods for solving job shop scheduling problems (JSSP) usually use off-the-shelf GNN models tailored to undirected graphs and neglect the rich and meaningful topological structures of disjunctive graphs (DGs). This paper proposes the topology-aware bidirectional graph attention network (TBGAT), a novel GNN architecture based on the attention mechanism, to embed the DG for solving JSSP in a local search framework. Specifically, TBGAT embeds the DG from a forward and a backward view, respectively, where the messages are propagated by following the different topologies of the views and aggregated via graph attention. Then, we propose a novel operator based on the message-passing mechanism to calculate the forward and backward topological sorts of the DG, which are the features for characterizing the topological structures and exploited by our model. In addition, we theoretically and experimentally show that TBGAT has linear computational complexity to the number of jobs and machines, respectively, strengthening our method's practical value. Besides, extensive experiments on five synthetic datasets and seven classic benchmarks show that TBGAT achieves new SOTA results by outperforming a wide range of neural methods by a large margin. All the code and data are publicly available online at https://github.com/zcaicaros/TBGAT.
Authors: Saurabh Mishra, Anant Raj, Sharan Vaswani
Abstract: Inverse optimization involves inferring unknown parameters of an optimization problem from known solutions and is widely used in fields such as transportation, power systems, and healthcare. We study the contextual inverse optimization setting that utilizes additional contextual information to better predict the unknown problem parameters. We focus on contextual inverse linear programming (CILP), addressing the challenges posed by the non-differentiable nature of LPs. For a linear prediction model, we reduce CILP to a convex feasibility problem allowing the use of standard algorithms such as alternating projections. The resulting algorithm for CILP is equipped with theoretical convergence guarantees without additional assumptions such as degeneracy or interpolation. Next, we reduce CILP to empirical risk minimization (ERM) on a smooth, convex loss that satisfies the Polyak-Lojasiewicz condition. This reduction enables the use of scalable first-order optimization methods to solve large non-convex problems while maintaining theoretical guarantees in the convex setting. Subsequently, we use the reduction to ERM to quantify the generalization performance of the proposed algorithm on previously unseen instances. Finally, we experimentally validate our approach on synthetic and real-world problems and demonstrate improved performance compared to existing methods.
Authors: Guangyi Liu, Yu Wang, Zeyu Feng, Qiyu Wu, Liping Tang, Yuan Gao, Zhen Li, Shuguang Cui, Julian McAuley, Zichao Yang, Eric P. Xing, Zhiting Hu
Abstract: The vast applications of deep generative models are anchored in three core capabilities -- generating new instances, reconstructing inputs, and learning compact representations -- across various data types, such as discrete text/protein sequences and continuous images. Existing model families, like variational autoencoders (VAEs), generative adversarial networks (GANs), autoregressive models, and (latent) diffusion models, generally excel in specific capabilities and data types but fall short in others. We introduce Generalized Encoding-Decoding Diffusion Probabilistic Models (EDDPMs) which integrate the core capabilities for broad applicability and enhanced performance. EDDPMs generalize the Gaussian noising-denoising in standard diffusion by introducing parameterized encoding-decoding. Crucially, EDDPMs are compatible with the well-established diffusion model objective and training recipes, allowing effective learning of the encoder-decoder parameters jointly with diffusion. By choosing appropriate encoder/decoder (e.g., large language models), EDDPMs naturally apply to different data types. Extensive experiments on text, proteins, and images demonstrate the flexibility to handle diverse data and tasks and the strong improvement over various existing models.
Authors: James Flemings, Murali Annavaram
Abstract: Large Language models (LLMs) are achieving state-of-the-art performance in many different downstream tasks. However, the increasing urgency of data privacy puts pressure on practitioners to train LLMs with Differential Privacy (DP) on private data. Concurrently, the exponential growth in parameter size of LLMs necessitates model compression before deployment of LLMs on resource-constrained devices or latency-sensitive applications. Differential privacy and model compression generally must trade off utility loss to achieve their objectives. Moreover, simultaneously applying both schemes can compound the utility degradation. To this end, we propose DistilDP: a novel differentially private knowledge distillation algorithm that exploits synthetic data generated by a differentially private teacher LLM. The knowledge of a teacher LLM is transferred onto the student in two ways: one way from the synthetic data itself -- the hard labels, and the other way by the output distribution of the teacher evaluated on the synthetic data -- the soft labels. Furthermore, if the teacher and student share a similar architectural structure, we can further distill knowledge by aligning the hidden representations between both. Our experimental results demonstrate that DistilDP can substantially improve the utility over existing baselines, at least $9.0$ PPL on the Big Patent dataset, with strong privacy parameters, $\epsilon=2$. These promising results progress privacy-preserving compression of autoregressive LLMs. Our code can be accessed here: https://github.com/james-flemings/dp_compress.
Authors: Shikun Liu, Deyu Zou, Han Zhao, Pan Li
Abstract: Graph-based methods, pivotal for label inference over interconnected objects in many real-world applications, often encounter generalization challenges, if the graph used for model training differs significantly from the graph used for testing. This work delves into Graph Domain Adaptation (GDA) to address the unique complexities of distribution shifts over graph data, where interconnected data points experience shifts in features, labels, and in particular, connecting patterns. We propose a novel, theoretically principled method, Pairwise Alignment (Pair-Align) to counter graph structure shift by mitigating conditional structure shift (CSS) and label shift (LS). Pair-Align uses edge weights to recalibrate the influence among neighboring nodes to handle CSS and adjusts the classification loss with label weights to handle LS. Our method demonstrates superior performance in real-world applications, including node classification with region shift in social networks, and the pileup mitigation task in particle colliding experiments. For the first application, we also curate the largest dataset by far for GDA studies. Our method shows strong performance in synthetic and other existing benchmark datasets.
Authors: Andi Nika, Debmalya Mandal, Parameswaran Kamalaruban, Georgios Tzannetos, Goran Radanovi\'c, Adish Singla
Abstract: In this paper, we take a step towards a deeper understanding of learning from human preferences by systematically comparing the paradigm of reinforcement learning from human feedback (RLHF) with the recently proposed paradigm of direct preference optimization (DPO). We focus our attention on the class of loglinear policy parametrization and linear reward functions. In order to compare the two paradigms, we first derive minimax statistical bounds on the suboptimality gap induced by both RLHF and DPO, assuming access to an oracle that exactly solves the optimization problems. We provide a detailed discussion on the relative comparison between the two paradigms, simultaneously taking into account the sample size, policy and reward class dimensions, and the regularization temperature. Moreover, we extend our analysis to the approximate optimization setting and derive exponentially decaying convergence rates for both RLHF and DPO. Next, we analyze the setting where the ground-truth reward is not realizable and find that, while RLHF incurs a constant additional error, DPO retains its asymptotically decaying gap by just tuning the temperature accordingly. Finally, we extend our comparison to the Markov decision process setting, where we generalize our results with exact optimization. To the best of our knowledge, we are the first to provide such a comparative analysis for RLHF and DPO.
Authors: Yuhao Wu, Jiangchao Yao, Xiaobo Xia, Jun Yu, Ruxin Wang, Bo Han, Tongliang Liu
Abstract: Despite the success of the carefully-annotated benchmarks, the effectiveness of existing graph neural networks (GNNs) can be considerably impaired in practice when the real-world graph data is noisily labeled. Previous explorations in sample selection have been demonstrated as an effective way for robust learning with noisy labels, however, the conventional studies focus on i.i.d data, and when moving to non-iid graph data and GNNs, two notable challenges remain: (1) nodes located near topological class boundaries are very informative for classification but cannot be successfully distinguished by the heuristic sample selection. (2) there is no available measure that considers the graph topological information to promote sample selection in a graph. To address this dilemma, we propose a $\textit{Topological Sample Selection}$ (TSS) method that boosts the informative sample selection process in a graph by utilising topological information. We theoretically prove that our procedure minimizes an upper bound of the expected risk under target clean distribution, and experimentally show the superiority of our method compared with state-of-the-art baselines.
Authors: Yu Huang, Zixin Wen, Yuejie Chi, Yingbin Liang
Abstract: Masked reconstruction, which predicts randomly masked patches from unmasked ones, has emerged as an important approach in self-supervised pretraining. However, the theoretical understanding of masked pretraining is rather limited, especially for the foundational architecture of transformers. In this paper, to the best of our knowledge, we provide the first end-to-end theoretical guarantee of learning one-layer transformers in masked reconstruction self-supervised pretraining. On the conceptual side, we posit a mechanism of how transformers trained with masked vision pretraining objectives produce empirically observed local and diverse attention patterns, on data distributions with spatial structures that highlight feature-position correlations. On the technical side, our end-to-end characterization of training dynamics in softmax-attention models simultaneously accounts for input and position embeddings, which is developed based on a careful analysis tracking the interplay between feature-wise and position-wise attention correlations.
Authors: Lingjiao Chen, Jared Quincy Davis, Boris Hanin, Peter Bailis, Ion Stoica, Matei Zaharia, James Zou
Abstract: Many recent state-of-the-art results in language tasks were achieved using compound systems that perform multiple Language Model (LM) calls and aggregate their responses. However, there is little understanding of how the number of LM calls - e.g., when asking the LM to answer each question multiple times and taking a majority vote - affects such a compound system's performance. In this paper, we initiate the study of scaling properties of compound inference systems. We analyze, theoretically and empirically, how the number of LM calls affects the performance of Vote and Filter-Vote, two of the simplest compound system designs, which aggregate LM responses via majority voting, optionally applying LM filters. We find, surprisingly, that across multiple language tasks, the performance of both Vote and Filter-Vote can first increase but then decrease as a function of the number of LM calls. Our theoretical results suggest that this non-monotonicity is due to the diversity of query difficulties within a task: more LM calls lead to higher performance on "easy" queries, but lower performance on "hard" queries, and non-monotone behavior can emerge when a task contains both types of queries. This insight then allows us to compute, from a small number of samples, the number of LM calls that maximizes system performance, and define an analytical scaling model for both systems. Experiments show that our scaling model can accurately predict the performance of Vote and Filter-Vote systems and thus find the optimal number of LM calls to make.
Authors: Sayantan Choudhury, Nazarii Tupitsa, Nicolas Loizou, Samuel Horvath, Martin Takac, Eduard Gorbunov
Abstract: Adaptive methods are extremely popular in machine learning as they make learning rate tuning less expensive. This paper introduces a novel optimization algorithm named KATE, which presents a scale-invariant adaptation of the well-known AdaGrad algorithm. We prove the scale-invariance of KATE for the case of Generalized Linear Models. Moreover, for general smooth non-convex problems, we establish a convergence rate of $O \left(\frac{\log T}{\sqrt{T}} \right)$ for KATE, matching the best-known ones for AdaGrad and Adam. We also compare KATE to other state-of-the-art adaptive algorithms Adam and AdaGrad in numerical experiments with different problems, including complex machine learning tasks like image classification and text classification on real data. The results indicate that KATE consistently outperforms AdaGrad and matches/surpasses the performance of Adam in all considered scenarios.
Authors: Hoang Phan, Andrew Gordon Wilson, Qi Lei
Abstract: Models trained on data composed of different groups or domains can suffer from severe performance degradation under distribution shifts. While recent methods have largely focused on optimizing the worst-group objective, this often comes at the expense of good performance on other groups. To address this problem, we introduce an optimization scheme to achieve good performance across groups and find a good solution for all without severely sacrificing performance on any of them. However, directly applying such optimization involves updating the parameters of the entire network, making it both computationally expensive and challenging. Thus, we introduce Controllable Prompt Tuning (CPT), which couples our approach with prompt-tuning techniques. On spurious correlation benchmarks, our procedures achieve state-of-the-art results across both transformer and non-transformer architectures, as well as unimodal and multimodal data, while requiring only 0.4% tunable parameters.
Authors: Philip Amortila, Dylan J. Foster, Akshay Krishnamurthy
Abstract: Exploration is a major challenge in reinforcement learning, especially for high-dimensional domains that require function approximation. We propose exploration objectives -- policy optimization objectives that enable downstream maximization of any reward function -- as a conceptual framework to systematize the study of exploration. Within this framework, we introduce a new objective, $L_1$-Coverage, which generalizes previous exploration schemes and supports three fundamental desiderata: 1. Intrinsic complexity control. $L_1$-Coverage is associated with a structural parameter, $L_1$-Coverability, which reflects the intrinsic statistical difficulty of the underlying MDP, subsuming Block and Low-Rank MDPs. 2. Efficient planning. For a known MDP, optimizing $L_1$-Coverage efficiently reduces to standard policy optimization, allowing flexible integration with off-the-shelf methods such as policy gradient and Q-learning approaches. 3. Efficient exploration. $L_1$-Coverage enables the first computationally efficient model-based and model-free algorithms for online (reward-free or reward-driven) reinforcement learning in MDPs with low coverability. Empirically, we find that $L_1$-Coverage effectively drives off-the-shelf policy optimization algorithms to explore the state space.
Authors: He Zhang, Chang Liu, Zun Wang, Xinran Wei, Siyuan Liu, Nanning Zheng, Bin Shao, Tie-Yan Liu
Abstract: Predicting the mean-field Hamiltonian matrix in density functional theory is a fundamental formulation to leverage machine learning for solving molecular science problems. Yet, its applicability is limited by insufficient labeled data for training. In this work, we highlight that Hamiltonian prediction possesses a self-consistency principle, based on which we propose self-consistency training, an exact training method that does not require labeled data. It distinguishes the task from predicting other molecular properties by the following benefits: (1) it enables the model to be trained on a large amount of unlabeled data, hence addresses the data scarcity challenge and enhances generalization; (2) it is more efficient than running DFT to generate labels for supervised training, since it amortizes DFT calculation over a set of queries. We empirically demonstrate the better generalization in data-scarce and out-of-distribution scenarios, and the better efficiency over DFT labeling. These benefits push forward the applicability of Hamiltonian prediction to an ever-larger scale.
Authors: Yucen Wang, Shenghua Wan, Le Gan, Shuai Feng, De-Chuan Zhan
Abstract: Model-based methods have significantly contributed to distinguishing task-irrelevant distractors for visual control. However, prior research has primarily focused on heterogeneous distractors like noisy background videos, leaving homogeneous distractors that closely resemble controllable agents largely unexplored, which poses significant challenges to existing methods. To tackle this problem, we propose Implicit Action Generator (IAG) to learn the implicit actions of visual distractors, and present a new algorithm named implicit Action-informed Diverse visual Distractors Distinguisher (AD3), that leverages the action inferred by IAG to train separated world models. Implicit actions effectively capture the behavior of background distractors, aiding in distinguishing the task-irrelevant components, and the agent can optimize the policy within the task-relevant state space. Our method achieves superior performance on various visual control tasks featuring both heterogeneous and homogeneous distractors. The indispensable role of implicit actions learned by IAG is also empirically validated.
Authors: Matthew Zurek, Yudong Chen
Abstract: We study the sample complexity of learning an $\varepsilon$-optimal policy in an average-reward Markov decision process (MDP) under a generative model. For weakly communicating MDPs, we establish the complexity bound $\widetilde{O}(SA\frac{H}{\varepsilon^2} )$, where $H$ is the span of the bias function of the optimal policy and $SA$ is the cardinality of the state-action space. Our result is the first that is minimax optimal (up to log factors) in all parameters $S,A,H$, and $\varepsilon$, improving on existing work that either assumes uniformly bounded mixing times for all policies or has suboptimal dependence on the parameters. We also initiate the study of sample complexity in general (multichain) average-reward MDPs. We argue a new transient time parameter $B$ is necessary, establish an $\widetilde{O}(SA\frac{B + H}{\varepsilon^2})$ complexity bound, and prove a matching (up to log factors) minimax lower bound. Both results are based on reducing the average-reward MDP to a discounted MDP, which requires new ideas in the general setting. To optimally analyze this reduction, we develop improved bounds for $\gamma$-discounted MDPs, showing that $\widetilde{O}(SA\frac{H}{(1-\gamma)^2\varepsilon^2} )$ and $\widetilde{O}(SA\frac{B + H}{(1-\gamma)^2\varepsilon^2} )$ samples suffice to learn $\varepsilon$-optimal policies in weakly communicating and in general MDPs, respectively. Both these results circumvent the well-known minimax lower bound of $\widetilde{\Omega}(SA\frac{1}{(1-\gamma)^3\varepsilon^2} )$ for $\gamma$-discounted MDPs, and establish a quadratic rather than cubic horizon dependence for a fixed MDP instance.
Authors: Jannis Chemseddine, Paul Hagemann, Gabriele Steidl, Christian Wald
Abstract: In inverse problems, many conditional generative models approximate the posterior measure by minimizing a distance between the joint measure and its learned approximation. While this approach also controls the distance between the posterior measures in the case of the Kullback--Leibler divergence, this is in general not hold true for the Wasserstein distance. In this paper, we introduce a conditional Wasserstein distance via a set of restricted couplings that equals the expected Wasserstein distance of the posteriors. Interestingly, the dual formulation of the conditional Wasserstein-1 flow resembles losses in the conditional Wasserstein GAN literature in a quite natural way. We derive theoretical properties of the conditional Wasserstein distance, characterize the corresponding geodesics and velocity fields as well as the flow ODEs. Subsequently, we propose to approximate the velocity fields by relaxing the conditional Wasserstein distance. Based on this, we propose an extension of OT Flow Matching for solving Bayesian inverse problems and demonstrate its numerical advantages on an inverse problem and class-conditional image generation.
Authors: Hyunki Seong, David Hyunchul Shim
Abstract: We introduce MoNet, a novel functionally modular network for self-supervised and interpretable end-to-end learning. By leveraging its functional modularity with a latent-guided contrastive loss function, MoNet efficiently learns task-specific decision-making processes in latent space without requiring task-level supervision. Moreover, our method incorporates an online, post-hoc explainability approach that enhances the interpretability of end-to-end inferences without compromising sensorimotor control performance. In real-world indoor environments, MoNet demonstrates effective visual autonomous navigation, outperforming baseline models by 7% to 28% in task specificity analysis. We further explore the interpretability of our network through post-hoc analysis of perceptual saliency maps and latent decision vectors. This provides valuable insights into the incorporation of explainable artificial intelligence into robotic learning, encompassing both perceptual and behavioral perspectives. Supplementary materials are available at https://sites.google.com/view/monet-lgc.
Authors: Neil Band, Xuechen Li, Tengyu Ma, Tatsunori Hashimoto
Abstract: Language models (LMs) may lead their users to make suboptimal downstream decisions when they confidently hallucinate. This issue can be mitigated by having the LM verbally convey the probability that its claims are correct, but existing models cannot produce long-form text with calibrated confidence statements. Through the lens of decision-making, we define linguistic calibration for long-form generations: an LM is linguistically calibrated if its generations enable its users to make calibrated probabilistic predictions. This definition enables a training framework where a supervised finetuning step bootstraps an LM to emit long-form generations with confidence statements such as "I estimate a 30% chance of..." or "I am certain that...", followed by a reinforcement learning step which rewards generations that enable a user to provide calibrated answers to related questions. We linguistically calibrate Llama 2 7B and find in automated and human evaluations of long-form generations that it is significantly more calibrated than strong finetuned factuality baselines with comparable accuracy. These findings generalize under significant domain shifts to scientific and biomedical questions and to an entirely held-out person biography generation task. Our results demonstrate that long-form generations may be calibrated end-to-end by constructing an objective in the space of the predictions that users make in downstream decision-making.
Authors: Amit Dhurandhar, Tejaswini Pedapati, Ronny Luss, Soham Dan, Aurelie Lozano, Payel Das, Georgios Kollias
Abstract: Transformer-based Language Models have become ubiquitous in Natural Language Processing (NLP) due to their impressive performance on various tasks. However, expensive training as well as inference remains a significant impediment to their widespread applicability. While enforcing sparsity at various levels of the model architecture has found promise in addressing scaling and efficiency issues, there remains a disconnect between how sparsity affects network topology. Inspired by brain neuronal networks, we explore sparsity approaches through the lens of network topology. Specifically, we exploit mechanisms seen in biological networks, such as preferential attachment and redundant synapse pruning, and show that principled, model-agnostic sparsity approaches are performant and efficient across diverse NLP tasks, spanning both classification (such as natural language inference) and generation (summarization, machine translation), despite our sole objective not being optimizing performance. NeuroPrune is competitive with (or sometimes superior to) baselines on performance and can be up to $10$x faster in terms of training time for a given level of sparsity, simultaneously exhibiting measurable improvements in inference time in many cases.
Authors: Lujie Yang, Hongkai Dai, Zhouxing Shi, Cho-Jui Hsieh, Russ Tedrake, Huan Zhang
Abstract: Learning-based neural network (NN) control policies have shown impressive empirical performance in a wide range of tasks in robotics and control. However, formal (Lyapunov) stability guarantees over the region-of-attraction (ROA) for NN controllers with nonlinear dynamical systems are challenging to obtain, and most existing approaches rely on expensive solvers such as sums-of-squares (SOS), mixed-integer programming (MIP), or satisfiability modulo theories (SMT). In this paper, we demonstrate a new framework for learning NN controllers together with Lyapunov certificates using fast empirical falsification and strategic regularizations. We propose a novel formulation that defines a larger verifiable region-of-attraction (ROA) than shown in the literature, and refines the conventional restrictive constraints on Lyapunov derivatives to focus only on certifiable ROAs. The Lyapunov condition is rigorously verified post-hoc using branch-and-bound with scalable linear bound propagation-based NN verification techniques. The approach is efficient and flexible, and the full training and verification procedure is accelerated on GPUs without relying on expensive solvers for SOS, MIP, nor SMT. The flexibility and efficiency of our framework allow us to demonstrate Lyapunov-stable output feedback control with synthesized NN-based controllers and NN-based observers with formal stability guarantees, for the first time in literature. Source code at https://github.com/Verified-Intelligence/Lyapunov_Stable_NN_Controllers
URLs: https://github.com/Verified-Intelligence/Lyapunov_Stable_NN_Controllers
Authors: William Merrill, Jackson Petty, Ashish Sabharwal
Abstract: State-space models (SSMs) have emerged as a potential alternative architecture for building large language models (LLMs) compared to the previously ubiquitous transformer architecture. One theoretical weakness of transformers is that they cannot express certain kinds of sequential computation and state tracking (Merrill & Sabharwal, 2023), which SSMs are explicitly designed to address via their close architectural similarity to recurrent neural networks (RNNs). But do SSMs truly have an advantage (over transformers) in expressive power for state tracking? Surprisingly, the answer is no. Our analysis reveals that the expressive power of SSMs is limited very similarly to transformers: SSMs cannot express computation outside the complexity class $\mathsf{TC}^0$. In particular, this means they cannot solve simple state-tracking problems like permutation composition. It follows that SSMs are provably unable to accurately track chess moves with certain notation, evaluate code, or track entities in a long narrative. To supplement our formal analysis, we report experiments showing that Mamba-style SSMs indeed struggle with state tracking. Thus, despite its recurrent formulation, the "state" in an SSM is an illusion: SSMs have similar expressiveness limitations to non-recurrent models like transformers, which may fundamentally limit their ability to solve real-world state-tracking problems.
Authors: Liheng Ma, Soumyasundar Pal, Yitian Zhang, Jiaming Zhou, Yingxue Zhang, Mark Coates
Abstract: The existing definitions of graph convolution, either from spatial or spectral perspectives, are inflexible and not unified. Defining a general convolution operator in the graph domain is challenging due to the lack of canonical coordinates, the presence of irregular structures, and the properties of graph symmetries. In this work, we propose a novel and general graph convolution framework by parameterizing the kernels as continuous functions of pseudo-coordinates derived via graph positional encoding. We name this Continuous Kernel Graph Convolution (CKGConv). Theoretically, we demonstrate that CKGConv is flexible and expressive. CKGConv encompasses many existing graph convolutions, and exhibits a stronger expressiveness, as powerful as graph transformers in terms of distinguishing non-isomorphic graphs. Empirically, we show that CKGConv-based Networks outperform existing graph convolutional networks and perform comparably to the best graph transformers across a variety of graph datasets. The code and models are publicly available at https://github.com/networkslab/CKGConv.
Authors: Victor Qu\'etu, Zhu Liao, Enzo Tartaglione
Abstract: While deep neural networks are highly effective at solving complex tasks, large pre-trained models are commonly employed even to solve consistently simpler downstream tasks, which do not necessarily require a large model's complexity. Motivated by the awareness of the ever-growing AI environmental impact, we propose an efficiency strategy that leverages prior knowledge transferred by large models. Simple but effective, we propose a method relying on an Entropy-bASed Importance mEtRic (EASIER) to reduce the depth of over-parametrized deep neural networks, which alleviates their computational burden. We assess the effectiveness of our method on traditional image classification setups. Our code is available at https://github.com/VGCQ/EASIER.
Authors: Ben Harwood, Amir Dezfouli, Iadine Chades, Conrad Sanderson
Abstract: Approximate k-Nearest Neighbour (ANN) methods are often used for mining information and aiding machine learning on large scale high-dimensional datasets. ANN methods typically differ in the index structure used for accelerating searches, resulting in various recall/runtime trade-off points. For applications with static datasets, runtime constraints and dataset properties can be used to empirically select an ANN method with suitable operating characteristics. However, for applications with dynamic datasets, which are subject to frequent online changes (like addition of new samples), there is currently no consensus as to which ANN methods are most suitable. Traditional evaluation approaches do not consider the computational costs of updating the index structure, as well as the rate and size of index updates. To address this, we empirically evaluate 5 popular ANN methods on two main applications (online data collection and online feature learning) while taking into account these considerations. Two dynamic datasets are used, derived from the SIFT1M dataset with 1 million samples and the DEEP1B dataset with 1 billion samples. The results indicate that the often used k-d trees method is not suitable on dynamic datasets as it is slower than a straightforward baseline exhaustive search method. For online data collection, the Hierarchical Navigable Small World Graphs method achieves a consistent speedup over baseline across a wide range of recall rates. For online feature learning, the Scalable Nearest Neighbours method is faster than baseline for recall rates below 75%.
Authors: Simon Vary, Pierre Ablin, Bin Gao, P. -A. Absil
Abstract: Optimization over the set of matrices $X$ that satisfy $X^\top B X = I_p$, referred to as the generalized Stiefel manifold, appears in many applications involving sampled covariance matrices such as the canonical correlation analysis (CCA), independent component analysis (ICA), and the generalized eigenvalue problem (GEVP). Solving these problems is typically done by iterative methods that require a fully formed $B$. We propose a cheap stochastic iterative method that solves the optimization problem while having access only to a random estimates of $B$. Our method does not enforce the constraint in every iteration; instead, it produces iterations that converge to critical points on the generalized Stiefel manifold defined in expectation. The method has lower per-iteration cost, requires only matrix multiplications, and has the same convergence rates as its Riemannian optimization counterparts that require the full matrix $B$. Experiments demonstrate its effectiveness in various machine learning applications involving generalized orthogonality constraints, including CCA, ICA, and the GEVP.
Authors: M. Saquib Sarfraz, Mei-Yen Chen, Lukas Layer, Kunyu Peng, Marios Koulakis
Abstract: The current state of machine learning scholarship in Timeseries Anomaly Detection (TAD) is plagued by the persistent use of flawed evaluation metrics, inconsistent benchmarking practices, and a lack of proper justification for the choices made in novel deep learning-based model designs. Our paper presents a critical analysis of the status quo in TAD, revealing the misleading track of current research and highlighting problematic methods, and evaluation practices. Our position advocates for a shift in focus from solely pursuing novel model designs to improving benchmarking practices, creating non-trivial datasets, and critically evaluating the utility of complex methods against simpler baselines. Our findings demonstrate the need for rigorous evaluation protocols, the creation of simple baselines, and the revelation that state-of-the-art deep anomaly detection models effectively learn linear mappings. These findings suggest the need for more exploration and development of simple and interpretable TAD methods. The increment of model complexity in the state-of-the-art deep-learning based models unfortunately offers very little improvement. We offer insights and suggestions for the field to move forward. Code: https://github.com/ssarfraz/QuoVadisTAD
Authors: Drew Prinster, Samuel Stanton, Anqi Liu, Suchi Saria
Abstract: As artificial intelligence (AI) / machine learning (ML) gain widespread adoption, practitioners are increasingly seeking means to quantify and control the risk these systems incur. This challenge is especially salient when such systems have autonomy to collect their own data, such as in black-box optimization and active learning, where their actions induce sequential feedback-loop shifts in the data distribution. Conformal prediction is a promising approach to uncertainty and risk quantification, but prior variants' validity guarantees have assumed some form of ``quasi-exchangeability'' on the data distribution, thereby excluding many types of sequential shifts. In this paper we prove that conformal prediction can theoretically be extended to \textit{any} joint data distribution, not just exchangeable or quasi-exchangeable ones. Although the most general case is exceedingly impractical to compute, for concrete practical applications we outline a procedure for deriving specific conformal algorithms for any data distribution, and we use this procedure to derive tractable algorithms for a series of AI/ML-agent-induced covariate shifts. We evaluate the proposed algorithms empirically on synthetic black-box optimization and active learning tasks.
Authors: Remi Genet, Hugo Inzirillo
Abstract: Recurrent Neural Networks (RNNs) have revolutionized many areas of machine learning, particularly in natural language and data sequence processing. Long Short-Term Memory (LSTM) has demonstrated its ability to capture long-term dependencies in sequential data. Inspired by the Kolmogorov-Arnold Networks (KANs) a promising alternatives to Multi-Layer Perceptrons (MLPs), we proposed a new neural networks architecture inspired by KAN and the LSTM, the Temporal Kolomogorov-Arnold Networks (TKANs). TKANs combined the strenght of both networks, it is composed of Recurring Kolmogorov-Arnold Networks (RKANs) Layers embedding memory management. This innovation enables us to perform multi-step time series forecasting with enhanced accuracy and efficiency. By addressing the limitations of traditional models in handling complex sequential patterns, the TKAN architecture offers significant potential for advancements in fields requiring more than one step ahead forecasting.
Authors: Marcin Podhajski, Jan Dubi\'nski, Franziska Boenisch, Adam Dziedzic, Agnieszka Pregowska, Tomasz Michalak
Abstract: Graph Neural Networks (GNNs) are recognized as potent tools for processing real-world data organized in graph structures. Especially inductive GNNs, which enable the processing of graph-structured data without relying on predefined graph structures, are gaining importance in an increasingly wide variety of applications. As these networks demonstrate proficiency across a range of tasks, they become lucrative targets for model-stealing attacks where an adversary seeks to replicate the functionality of the targeted network. A large effort has been made to develop model-stealing attacks that focus on models trained with images and texts. However, little attention has been paid to GNNs trained on graph data. This paper introduces a novel method for unsupervised model-stealing attacks against inductive GNNs, based on graph contrasting learning and spectral graph augmentations to efficiently extract information from the target model. The proposed attack is thoroughly evaluated on six datasets. The results show that this approach demonstrates a higher level of efficiency compared to existing stealing attacks. More concretely, our attack outperforms the baseline on all benchmarks achieving higher fidelity and downstream accuracy of the stolen model while requiring fewer queries sent to the target model.
Authors: Tehila Dahan, Kfir Y. Levy
Abstract: In this paper, we investigate the challenging framework of Byzantine-robust training in distributed machine learning (ML) systems, focusing on enhancing both efficiency and practicality. As distributed ML systems become integral for complex ML tasks, ensuring resilience against Byzantine failures-where workers may contribute incorrect updates due to malice or error-gains paramount importance. Our first contribution is the introduction of the Centered Trimmed Meta Aggregator (CTMA), an efficient meta-aggregator that upgrades baseline aggregators to optimal performance levels, while requiring low computational demands. Additionally, we propose harnessing a recently developed gradient estimation technique based on a double-momentum strategy within the Byzantine context. Our paper highlights its theoretical and practical advantages for Byzantine-robust training, especially in simplifying the tuning process and reducing the reliance on numerous hyperparameters. The effectiveness of this technique is supported by theoretical insights within the stochastic convex optimization (SCO) framework and corroborated by empirical evidence.
Authors: Zhangkai Wu, Xuhui Fan, Jin Li, Zhilin Zhao, Hui Chen, Longbing Cao
Abstract: The recently proposed Bayesian Flow Networks~(BFNs) show great potential in modeling parameter spaces, offering a unified strategy for handling continuous, discretized, and discrete data. However, BFNs cannot learn high-level semantic representation from the parameter space since {common encoders, which encode data into one static representation, cannot capture semantic changes in parameters.} This motivates a new direction: learning semantic representations hidden in the parameter spaces to characterize mixed-typed noisy data. {Accordingly, we propose a representation learning framework named ParamReL, which operates in the parameter space to obtain parameter-wise latent semantics that exhibit progressive structures. Specifically, ParamReL proposes a \emph{self-}encoder to learn latent semantics directly from parameters, rather than from observations. The encoder is then integrated into BFNs, enabling representation learning with various formats of observations. Mutual information terms further promote the disentanglement of latent semantics and capture meaningful semantics simultaneously.} We illustrate {conditional generation and reconstruction} in ParamReL via expanding BFNs, and extensive {quantitative} experimental results demonstrate the {superior effectiveness} of ParamReL in learning parameter representation.
Authors: Penghui Qi, Xinyi Wan, Nyamdavaa Amar, Min Lin
Abstract: Pipeline parallelism has been widely explored, but most existing schedules lack a systematic methodology. In this paper, we propose a framework to decompose pipeline schedules as repeating a building block and we show that the lifespan of the building block decides the peak activation memory of the pipeline schedule. Guided by the observations, we find that almost all existing pipeline schedules, to the best of our knowledge, are memory inefficient. To address this, we introduce a family of memory efficient building blocks with controllable activation memory, which can reduce the peak activation memory to 1/2 of 1F1B without sacrificing efficiency, and even to 1/3 with comparable throughput. We can also achieve almost zero pipeline bubbles while maintaining the same activation memory as 1F1B. Our evaluations demonstrate that in pure pipeline parallelism settings, our methods outperform 1F1B by from 7% to 55% in terms of throughput. When employing a grid search over hybrid parallelism hyperparameters in practical scenarios, our proposed methods demonstrate a 16% throughput improvement over the 1F1B baseline for large language models.
Authors: Hancheng Min, Ren\'e Vidal
Abstract: The implicit bias of gradient-based training algorithms has been considered mostly beneficial as it leads to trained networks that often generalize well. However, Frei et al. (2023) show that such implicit bias can harm adversarial robustness. Specifically, they show that if the data consists of clusters with small inter-cluster correlation, a shallow (two-layer) ReLU network trained by gradient flow generalizes well, but it is not robust to adversarial attacks of small radius. Moreover, this phenomenon occurs despite the existence of a much more robust classifier that can be explicitly constructed from a shallow network. In this paper, we extend recent analyses of neuron alignment to show that a shallow network with a polynomial ReLU activation (pReLU) trained by gradient flow not only generalizes well but is also robust to adversarial attacks. Our results highlight the importance of the interplay between data structure and architecture design in the implicit bias and robustness of trained networks.
Authors: Xing Hu, Yuan Cheng, Dawei Yang, Zhihang Yuan, Jiangyong Yu, Chen Xu, Sifan Zhou
Abstract: Post-training quantization (PTQ) serves as a potent technique to accelerate the inference of large language models (LLMs). Nonetheless, existing works still necessitate a considerable number of floating-point (FP) operations during inference, including additional quantization and de-quantization, as well as non-linear operators such as RMSNorm and Softmax. This limitation hinders the deployment of LLMs on the edge and cloud devices. In this paper, we identify the primary obstacle to integer-only quantization for LLMs lies in the large fluctuation of activations across channels and tokens in both linear and non-linear operations. To address this issue, we propose I-LLM, a novel integer-only fully-quantized PTQ framework tailored for LLMs. Specifically, (1) we develop Fully-Smooth Block-Reconstruction (FSBR) to aggressively smooth inter-channel variations of all activations and weights. (2) to alleviate degradation caused by inter-token variations, we introduce a novel approach called Dynamic Integer-only MatMul (DI-MatMul). This method enables dynamic quantization in full-integer matrix multiplication by dynamically quantizing the input and outputs with integer-only operations. (3) we design DI-ClippedSoftmax, DI-Exp, and DI-Normalization, which utilize bit shift to execute non-linear operators efficiently while maintaining accuracy. The experiment shows that our I-LLM achieves comparable accuracy to the FP baseline and outperforms non-integer quantization methods. For example, I-LLM can operate at W4A4 with negligible loss of accuracy. To our knowledge, we are the first to bridge the gap between integer-only quantization and LLMs. We've published our code on anonymous.4open.science, aiming to contribute to the advancement of this field.
Authors: Shicong Cen, Jincheng Mei, Katayoon Goshvadi, Hanjun Dai, Tong Yang, Sherry Yang, Dale Schuurmans, Yuejie Chi, Bo Dai
Abstract: Reinforcement learning from human feedback (RLHF) has demonstrated great promise in aligning large language models (LLMs) with human preference. Depending on the availability of preference data, both online and offline RLHF are active areas of investigation. A key bottleneck is understanding how to incorporate uncertainty estimation in the reward function learned from the preference data for RLHF, regardless of how the preference data is collected. While the principles of optimism or pessimism under uncertainty are well-established in standard reinforcement learning (RL), a practically-implementable and theoretically-grounded form amenable to large language models is not yet available, as standard techniques for constructing confidence intervals become intractable under arbitrary policy parameterizations. In this paper, we introduce a unified approach to online and offline RLHF -- value-incentivized preference optimization (VPO) -- which regularizes the maximum-likelihood estimate of the reward function with the corresponding value function, modulated by a $\textit{sign}$ to indicate whether the optimism or pessimism is chosen. VPO also directly optimizes the policy with implicit reward modeling, and therefore shares a simpler RLHF pipeline similar to direct preference optimization. Theoretical guarantees of VPO are provided for both online and offline settings, matching the rates of their standard RL counterparts. Moreover, experiments on text summarization and dialog verify the practicality and effectiveness of VPO.
Authors: Hengkai Tan, Songming Liu, Kai Ma, Chengyang Ying, Xingxing Zhang, Hang Su, Jun Zhu
Abstract: Transformer has shown promise in reinforcement learning to model time-varying features for obtaining generalized low-level robot policies on diverse robotics datasets in embodied learning. However, it still suffers from the issues of low data efficiency and high inference latency. In this paper, we propose to investigate the task from a new perspective of the frequency domain. We first observe that the energy density in the frequency domain of a robot's trajectory is mainly concentrated in the low-frequency part. Then, we present the Fourier Controller Network (FCNet), a new network that uses Short-Time Fourier Transform (STFT) to extract and encode time-varying features through frequency domain interpolation. In order to do real-time decision-making, we further adopt FFT and Sliding DFT methods in the model architecture to achieve parallel training and efficient recurrent inference. Extensive results in both simulated (e.g., D4RL) and real-world environments (e.g., robot locomotion) demonstrate FCNet's substantial efficiency and effectiveness over existing methods such as Transformer, e.g., FCNet outperforms Transformer on multi-environmental robotics datasets of all types of sizes (from 1.9M to 120M). The project page and code can be found https://thkkk.github.io/fcnet.
Authors: Jaerin Lee, Bong Gyun Kang, Kihoon Kim, Kyoung Mu Lee
Abstract: One puzzling artifact in machine learning dubbed grokking is where delayed generalization is achieved tenfolds of iterations after near perfect overfitting to the training data. Focusing on the long delay itself on behalf of machine learning practitioners, our goal is to accelerate generalization of a model under grokking phenomenon. By regarding a series of gradients of a parameter over training iterations as a random signal over time, we can spectrally decompose the parameter trajectories under gradient descent into two components: the fast-varying, overfitting-yielding component and the slow-varying, generalization-inducing component. This analysis allows us to accelerate the grokking phenomenon more than $\times 50$ with only a few lines of code that amplifies the slow-varying components of gradients. The experiments show that our algorithm applies to diverse tasks involving images, languages, and graphs, enabling practical availability of this peculiar artifact of sudden generalization. Our code is available at https://github.com/ironjr/grokfast.
Authors: Gezheng Xu, Qi Chen, Charles Ling, Boyu Wang, Changjian Shui
Abstract: AI systems have been shown to produce unfair results for certain subgroups of population, highlighting the need to understand bias on certain sensitive attributes. Current research often falls short, primarily focusing on the subgroups characterized by a single sensitive attribute, while neglecting the nature of intersectional fairness of multiple sensitive attributes. This paper focuses on its one fundamental aspect by discovering diverse high-bias subgroups under intersectional sensitive attributes. Specifically, we propose a Bias-Guided Generative Network (BGGN). By treating each bias value as a reward, BGGN efficiently generates high-bias intersectional sensitive attributes. Experiments on real-world text and image datasets demonstrate a diverse and efficient discovery of BGGN. To further evaluate the generated unseen but possible unfair intersectional sensitive attributes, we formulate them as prompts and use modern generative AI to produce new texts and images. The results of frequently generating biased data provides new insights of discovering potential unfairness in popular modern generative AI systems. Warning: This paper contains generative examples that are offensive in nature.
Authors: Davide Paglieri, Saurabh Dash, Tim Rockt\"aschel, Jack Parker-Holder
Abstract: Post-Training Quantization (PTQ) enhances the efficiency of Large Language Models (LLMs) by enabling faster operation and compatibility with more accessible hardware through reduced memory usage, at the cost of small performance drops. We explore the role of calibration sets in PTQ, specifically their effect on hidden activations in various notable open-source LLMs. Calibration sets are crucial for evaluating activation magnitudes and identifying outliers, which can distort the quantization range and negatively impact performance. Our analysis reveals a marked contrast in quantization effectiveness across models. The older OPT model, upon which much of the quantization literature is based, shows significant performance deterioration and high susceptibility to outliers with varying calibration sets. In contrast, newer models like Llama-2 7B, Llama-3 8B, Command-R 35B, and Mistral 7B demonstrate strong robustness, with Mistral 7B showing near-immunity to outliers and stable activations. These findings suggest a shift in PTQ strategies might be needed. As advancements in pre-training methods reduce the relevance of outliers, there is an emerging need to reassess the fundamentals of current quantization literature. The emphasis should pivot towards optimizing inference speed, rather than primarily focusing on outlier preservation, to align with the evolving characteristics of state-of-the-art LLMs.
Authors: Xiaojun Jia, Tianyu Pang, Chao Du, Yihao Huang, Jindong Gu, Yang Liu, Xiaochun Cao, Min Lin
Abstract: Large language models (LLMs) are being rapidly developed, and a key component of their widespread deployment is their safety-related alignment. Many red-teaming efforts aim to jailbreak LLMs, where among these efforts, the Greedy Coordinate Gradient (GCG) attack's success has led to a growing interest in the study of optimization-based jailbreaking techniques. Although GCG is a significant milestone, its attacking efficiency remains unsatisfactory. In this paper, we present several improved (empirical) techniques for optimization-based jailbreaks like GCG. We first observe that the single target template of "Sure" largely limits the attacking performance of GCG; given this, we propose to apply diverse target templates containing harmful self-suggestion and/or guidance to mislead LLMs. Besides, from the optimization aspects, we propose an automatic multi-coordinate updating strategy in GCG (i.e., adaptively deciding how many tokens to replace in each step) to accelerate convergence, as well as tricks like easy-to-hard initialisation. Then, we combine these improved technologies to develop an efficient jailbreak method, dubbed I-GCG. In our experiments, we evaluate on a series of benchmarks (such as NeurIPS 2023 Red Teaming Track). The results demonstrate that our improved techniques can help GCG outperform state-of-the-art jailbreaking attacks and achieve nearly 100% attack success rate. The code is released at https://github.com/jiaxiaojunQAQ/I-GCG.
Authors: Yuwei Fu, Haichao Zhang, Di Wu, Wei Xu, Benoit Boulet
Abstract: In this work, we investigate how to leverage pre-trained visual-language models (VLM) for online Reinforcement Learning (RL). In particular, we focus on sparse reward tasks with pre-defined textual task descriptions. We first identify the problem of reward misalignment when applying VLM as a reward in RL tasks. To address this issue, we introduce a lightweight fine-tuning method, named Fuzzy VLM reward-aided RL (FuRL), based on reward alignment and relay RL. Specifically, we enhance the performance of SAC/DrQ baseline agents on sparse reward tasks by fine-tuning VLM representations and using relay RL to avoid local minima. Extensive experiments on the Meta-world benchmark tasks demonstrate the efficacy of the proposed method. Code is available at: https://github.com/fuyw/FuRL.
Authors: Sofie Goethals, Eoin Delaney, Brent Mittelstadt, Chris Russell
Abstract: Access to resources strongly constrains the decisions we make. While we might wish to offer every student a scholarship, or schedule every patient for follow-up meetings with a specialist, limited resources mean that this is not possible. Existing tools for fair machine learning ignore these key constraints, with the majority of methods disregarding any finite resource limitations under which decisions are made. Our research introduces the concept of "resource-constrained fairness" and quantifies the cost of fairness within this framework. We demonstrate that the level of available resources significantly influences this cost, a factor that has been overlooked in previous evaluations.
Authors: Eugene Choi, Arash Ahmadian, Matthieu Geist, Oilvier Pietquin, Mohammad Gheshlaghi Azar
Abstract: Both online and offline RLHF methods such as PPO and DPO have been extremely successful in aligning AI with human preferences. Despite their success, the existing methods suffer from a fundamental problem that their optimal solution is highly task-dependent (i.e., not robust to out-of-distribution (OOD) tasks). Here we address this challenge by proposing Self-Improving Robust Preference Optimization SRPO, a practical and mathematically principled offline RLHF framework that is completely robust to the changes in the task. The key idea of SRPO is to cast the problem of learning from human preferences as a self-improvement process, which can be mathematically expressed in terms of a min-max objective that aims at joint optimization of self-improvement policy and the generative policy in an adversarial fashion. The solution for this optimization problem is independent of the training task and thus it is robust to its changes. We then show that this objective can be re-expressed in the form of a non-adversarial offline loss which can be optimized using standard supervised optimization techniques at scale without any need for reward model and online inference. We show the effectiveness of SRPO in terms of AI Win-Rate (WR) against human (GOLD) completions. In particular, when SRPO is evaluated on the OOD XSUM dataset, it outperforms the celebrated DPO by a clear margin of 15% after 5 self-revisions, achieving WR of 90%.
Authors: Yunni Qu (Department of Computer Science, University of North Carolina at Chapel Hill), James Wellnitz (Eshelman School of Pharmacy, University of North Carolina at Chapel Hill), Alexander Tropsha (Eshelman School of Pharmacy, University of North Carolina at Chapel Hill), Junier Oliva (Department of Computer Science, University of North Carolina at Chapel Hill)
Abstract: Expansive Matching of Experts (EMOE) is a novel method that utilizes support-expanding, extrapolatory pseudo-labeling to improve prediction and uncertainty based rejection on out-of-distribution (OOD) points. We propose an expansive data augmentation technique that generates OOD instances in a latent space, and an empirical trial based approach to filter out augmented expansive points for pseudo-labeling. EMOE utilizes a diverse set of multiple base experts as pseudo-labelers on the augmented data to improve OOD performance through a shared MLP with multiple heads (one per expert). We demonstrate that EMOE achieves superior performance compared to state-of-the-art methods on tabular data.
Authors: Jianrong Ding, Zhanyu Liu, Guanjie Zheng, Haiming Jin, Linghe Kong
Abstract: Dataset condensation is a newborn technique that generates a small dataset that can be used in training deep neural networks to lower training costs. The objective of dataset condensation is to ensure that the model trained with the synthetic dataset can perform comparably to the model trained with full datasets. However, existing methods predominantly concentrate on classification tasks, posing challenges in their adaptation to time series forecasting (TS-forecasting). This challenge arises from disparities in the evaluation of synthetic data. In classification, the synthetic data is considered well-distilled if the model trained with the full dataset and the model trained with the synthetic dataset yield identical labels for the same input, regardless of variations in output logits distribution. Conversely, in TS-forecasting, the effectiveness of synthetic data distillation is determined by the distance between predictions of the two models. The synthetic data is deemed well-distilled only when all data points within the predictions are similar. Consequently, TS-forecasting has a more rigorous evaluation methodology compared to classification. To mitigate this gap, we theoretically analyze the optimization objective of dataset condensation for TS-forecasting and propose a new one-line plugin of dataset condensation designated as Dataset Condensation for Time Series Forecasting (CondTSF) based on our analysis. Plugging CondTSF into previous dataset condensation methods facilitates a reduction in the distance between the predictions of the model trained with the full dataset and the model trained with the synthetic dataset, thereby enhancing performance. We conduct extensive experiments on eight commonly used time series datasets. CondTSF consistently improves the performance of all previous dataset condensation methods across all datasets, particularly at low condensing ratios.
Authors: Louis Serrano, Thomas X Wang, Etienne Le Naour, Jean-No\"el Vittaut, Patrick Gallinari
Abstract: We present AROMA (Attentive Reduced Order Model with Attention), a framework designed to enhance the modeling of partial differential equations (PDEs) using local neural fields. Our flexible encoder-decoder architecture can obtain smooth latent representations of spatial physical fields from a variety of data types, including irregular-grid inputs and point clouds. This versatility eliminates the need for patching and allows efficient processing of diverse geometries. The sequential nature of our latent representation can be interpreted spatially and permits the use of a conditional transformer for modeling the temporal dynamics of PDEs. By employing a diffusion-based formulation, we achieve greater stability and enable longer rollouts compared to conventional MSE training. AROMA's superior performance in simulating 1D and 2D equations underscores the efficacy of our approach in capturing complex dynamical behaviors.
Authors: Youwei Shu, Xi Xiao, Derui Wang, Yuxin Cao, Siji Chen, Jason Xue, Linyi Li, Bo Li
Abstract: Randomized Smoothing (RS) is currently a scalable certified defense method providing robustness certification against adversarial examples. Although significant progress has been achieved in providing defenses against $\ell_p$ adversaries, the interaction between the smoothing distribution and the robustness certification still remains vague. In this work, we comprehensively study the effect of two families of distributions, named Exponential Standard Gaussian (ESG) and Exponential General Gaussian (EGG) distributions, on Randomized Smoothing and Double Sampling Randomized Smoothing (DSRS). We derive an analytic formula for ESG's certified radius, which converges to the origin formula of RS as the dimension $d$ increases. Additionally, we prove that EGG can provide tighter constant factors than DSRS in providing $\Omega(\sqrt{d})$ lower bounds of $\ell_2$ certified radius, and thus further addresses the curse of dimensionality in RS. Our experiments on real-world datasets confirm our theoretical analysis of the ESG distributions, that they provide almost the same certification under different exponents $\eta$ for both RS and DSRS. In addition, EGG brings a significant improvement to the DSRS certification, but the mechanism can be different when the classifier properties are different. Compared to the primitive DSRS, the increase in certified accuracy provided by EGG is prominent, up to 6.4% on ImageNet.
Authors: Kathrin Donandt, Dirk S\"offker
Abstract: Using data sources beyond the Automatic Identification System to represent the context a vessel is navigating in and consequently improve situation awareness is still rare in machine learning approaches to vessel trajectory prediction (VTP). In inland shipping, where vessel movement is constrained within fairways, navigational context information is indispensable. In this contribution targeting inland VTP, Gaussian Mixture Models (GMMs) are applied, on a fused dataset of AIS and discharge measurements, to generate multi-modal distribution curves, capturing typical lateral vessel positioning in the fairway and dislocation speeds along the waterway. By sampling the probability density curves of the GMMs, feature vectors are derived which are used, together with spatio-temporal vessel features and fairway geometries, as input to a VTP transformer model. The incorporation of these distribution features of both the current and forthcoming navigation context improves prediction accuracy. The superiority of the model over a previously proposed transformer model for inland VTP is shown. The novelty lies in the provision of preprocessed, statistics-based features representing the conditioned spatial context, rather than relying on the model to extract relevant features for the VTP task from contextual data. Oversimplification of the complexity of inland navigation patterns by assuming a single typical route or selecting specific clusters prior to model application is avoided by giving the model access to the entire distribution information. The methodology's generalizability is demonstrated through the usage of data of 3 distinct river sections. It can be integrated into an interaction-aware prediction framework, where insights into the positioning of the actual vessel behavior in the overall distribution at the current location and discharge can enhance trajectory prediction accuracy.
Authors: Katarina Petrovi\'c, Shenyang Huang, Farimah Poursafaei, Petar Veli\v{c}kovi\'c
Abstract: Evolving relations in real-world networks are often modelled by temporal graphs. Graph rewiring techniques have been utilised on Graph Neural Networks (GNNs) to improve expressiveness and increase model performance. In this work, we propose Temporal Graph Rewiring (TGR), the first approach for graph rewiring on temporal graphs. TGR enables communication between temporally distant nodes in a continuous time dynamic graph by utilising expander graph propagation to construct a message passing highway for message passing between distant nodes. Expander graphs are suitable candidates for rewiring as they help overcome the oversquashing problem often observed in GNNs. On the public tgbl-wiki benchmark, we show that TGR improves the performance of a widely used TGN model by a significant margin. Our code repository is accessible at https://github.com/kpetrovicc/TGR.git .
Authors: Remi Genet, Hugo Inzirillo
Abstract: Capturing complex temporal patterns and relationships within multivariate data streams is a difficult task. We propose the Temporal Kolmogorov-Arnold Transformer (TKAT), a novel attention-based architecture designed to address this task using Temporal Kolmogorov-Arnold Networks (TKANs). Inspired by the Temporal Fusion Transformer (TFT), TKAT emerges as a powerful encoder-decoder model tailored to handle tasks in which the observed part of the features is more important than the a priori known part. This new architecture combined the theoretical foundation of the Kolmogorov-Arnold representation with the power of transformers. TKAT aims to simplify the complex dependencies inherent in time series, making them more "interpretable". The use of transformer architecture in this framework allows us to capture long-range dependencies through self-attention mechanisms.
Authors: Haibin Wu, Xu Li, Andy T. Liu, Zhiyong Wu, Helen Meng, Hung-yi Lee
Abstract: Previous works have shown that automatic speaker verification (ASV) is seriously vulnerable to malicious spoofing attacks, such as replay, synthetic speech, and recently emerged adversarial attacks. Great efforts have been dedicated to defending ASV against replay and synthetic speech; however, only a few approaches have been explored to deal with adversarial attacks. All the existing approaches to tackle adversarial attacks for ASV require the knowledge for adversarial samples generation, but it is impractical for defenders to know the exact attack algorithms that are applied by the in-the-wild attackers. This work is among the first to perform adversarial defense for ASV without knowing the specific attack algorithms. Inspired by self-supervised learning models (SSLMs) that possess the merits of alleviating the superficial noise in the inputs and reconstructing clean samples from the interrupted ones, this work regards adversarial perturbations as one kind of noise and conducts adversarial defense for ASV by SSLMs. Specifically, we propose to perform adversarial defense from two perspectives: 1) adversarial perturbation purification and 2) adversarial perturbation detection. Experimental results show that our detection module effectively shields the ASV by detecting adversarial samples with an accuracy of around 80%. Moreover, since there is no common metric for evaluating the adversarial defense performance for ASV, this work also formalizes evaluation metrics for adversarial defense considering both purification and detection based approaches into account. We sincerely encourage future works to benchmark their approaches based on the proposed evaluation framework.
Authors: Jitendra K. Tugnait
Abstract: We consider the problem of inferring the conditional independence graph (CIG) of a sparse, high-dimensional stationary multivariate Gaussian time series. A sparse-group lasso-based frequency-domain formulation of the problem based on frequency-domain sufficient statistic for the observed time series is presented. We investigate an alternating direction method of multipliers (ADMM) approach for optimization of the sparse-group lasso penalized log-likelihood. We provide sufficient conditions for convergence in the Frobenius norm of the inverse PSD estimators to the true value, jointly across all frequencies, where the number of frequencies are allowed to increase with sample size. This results also yields a rate of convergence. We also empirically investigate selection of the tuning parameters based on Bayesian information criterion, and illustrate our approach using numerical examples utilizing both synthetic and real data.
Authors: Luhao Zhang, Jincheng Yang, Rui Gao
Abstract: We present a general duality result for Wasserstein distributionally robust optimization that holds for any Kantorovich transport cost, measurable loss function, and nominal probability distribution. Assuming an interchangeability principle inherent in existing duality results, our proof only uses one-dimensional convex analysis. Furthermore, we demonstrate that the interchangeability principle holds if and only if certain measurable projection and weak measurable selection conditions are satisfied. To illustrate the broader applicability of our approach, we provide a rigorous treatment of duality results in distributionally robust Markov decision processes and distributionally robust multistage stochastic programming. Additionally, we extend our analysis to other problems such as infinity-Wasserstein distributionally robust optimization, risk-averse optimization, and globalized distributionally robust counterpart.
Authors: Chen Cheng, Hilal Asi, John Duchi
Abstract: The construction of most supervised learning datasets revolves around collecting multiple labels for each instance, then aggregating the labels to form a type of "gold-standard". We question the wisdom of this pipeline by developing a (stylized) theoretical model of this process and analyzing its statistical consequences, showing how access to non-aggregated label information can make training well-calibrated models more feasible than it is with gold-standard labels. The entire story, however, is subtle, and the contrasts between aggregated and fuller label information depend on the particulars of the problem, where estimators that use aggregated information exhibit robust but slower rates of convergence, while estimators that can effectively leverage all labels converge more quickly if they have fidelity to (or can learn) the true labeling process. The theory makes several predictions for real-world datasets, including when non-aggregate labels should improve learning performance, which we test to corroborate the validity of our predictions.
Authors: Khoa D. Doan, Jianwen Xie, Yaxuan Zhu, Yang Zhao, Ping Li
Abstract: Leveraging supervised information can lead to superior retrieval performance in the image hashing domain but the performance degrades significantly without enough labeled data. One effective solution to boost performance is to employ generative models, such as Generative Adversarial Networks (GANs), to generate synthetic data in an image hashing model. However, GAN-based methods are difficult to train, which prevents the hashing approaches from jointly training the generative models and the hash functions. This limitation results in sub-optimal retrieval performance. To overcome this limitation, we propose a novel framework, the generative cooperative hashing network, which is based on energy-based cooperative learning. This framework jointly learns a powerful generative representation of the data and a robust hash function via two components: a top-down contrastive pair generator that synthesizes contrastive images and a bottom-up multipurpose descriptor that simultaneously represents the images from multiple perspectives, including probability density, hash code, latent code, and category. The two components are jointly learned via a novel likelihood-based cooperative learning scheme. We conduct experiments on several real-world datasets and show that the proposed method outperforms the competing hashing supervised methods, achieving up to 10\% relative improvement over the current state-of-the-art supervised hashing methods, and exhibits a significantly better performance in out-of-distribution retrieval.
Authors: Kajsa M{\o}llersen, Einar Holsb{\o}
Abstract: Machine learning methods are commonly evaluated and compared by their performance on data sets from public repositories. This allows for multiple methods, oftentimes several thousands, to be evaluated under identical conditions and across time. The highest ranked performance on a problem is referred to as state-of-the-art (SOTA) performance, and is used, among other things, as a reference point for publication of new methods. Using the highest-ranked performance as an estimate for SOTA is a biased estimator, giving overly optimistic results. The mechanisms at play are those of multiplicity, a topic that is well-studied in the context of multiple comparisons and multiple testing, but has, as far as the authors are aware of, been nearly absent from the discussion regarding SOTA estimates. The optimistic state-of-the-art estimate is used as a standard for evaluating new methods, and methods with substantial inferior results are easily overlooked. In this article, we provide a probability distribution for the case of multiple classifiers so that known analyses methods can be engaged and a better SOTA estimate can be provided. We demonstrate the impact of multiplicity through a simulated example with independent classifiers. We show how classifier dependency impacts the variance, but also that the impact is limited when the accuracy is high. Finally, we discuss three real-world examples; Kaggle competitions that demonstrate various aspects.
Authors: Jorge Lor\'ia, Anindya Bhadra
Abstract: From the classical and influential works of Neal (1996), it is known that the infinite width scaling limit of a Bayesian neural network with one hidden layer is a Gaussian process, when the network weights have bounded prior variance. Neal's result has been extended to networks with multiple hidden layers and to convolutional neural networks, also with Gaussian process scaling limits. The tractable properties of Gaussian processes then allow straightforward posterior inference and uncertainty quantification, considerably simplifying the study of the limit process compared to a network of finite width. Neural network weights with unbounded variance, however, pose unique challenges. In this case, the classical central limit theorem breaks down and it is well known that the scaling limit is an $\alpha$-stable process under suitable conditions. However, current literature is primarily limited to forward simulations under these processes and the problem of posterior inference under such a scaling limit remains largely unaddressed, unlike in the Gaussian process case. To this end, our contribution is an interpretable and computationally efficient procedure for posterior inference, using a conditionally Gaussian representation, that then allows full use of the Gaussian process machinery for tractable posterior inference and uncertainty quantification in the non-Gaussian regime.
Authors: Ziyu Chen, Markos A. Katsoulakis, Luc Rey-Bellet, Wei Zhu
Abstract: Group-invariant generative adversarial networks (GANs) are a type of GANs in which the generators and discriminators are hardwired with group symmetries. Empirical studies have shown that these networks are capable of learning group-invariant distributions with significantly improved data efficiency. In this study, we aim to rigorously quantify this improvement by analyzing the reduction in sample complexity for group-invariant GANs. Our findings indicate that when learning group-invariant distributions, the number of samples required for group-invariant GANs decreases proportionally by a factor of the group size. Importantly, this sample complexity reduction cannot be achieved merely through data augmentation due to the probabilistic dependence of augmented data. Numerical results substantiate our theory and highlight the stark contrast between learning with group-invariant GANs and using data augmentation. This work presents the first statistical performance guarantees for group-invariant generative models, specifically for GANs, and it may shed light on the study of other generative models with group symmetries.
Authors: Kenshi Abe, Kaito Ariu, Mitsuki Sakamoto, Atsushi Iwasaki
Abstract: This paper proposes a payoff perturbation technique for the Mirror Descent (MD) algorithm in games where the gradient of the payoff functions is monotone in the strategy profile space, potentially containing additive noise. The optimistic family of learning algorithms, exemplified by optimistic MD, successfully achieves {\it last-iterate} convergence in scenarios devoid of noise, leading the dynamics to a Nash equilibrium. A recent re-emerging trend underscores the promise of the perturbation approach, where payoff functions are perturbed based on the distance from an anchoring, or {\it slingshot}, strategy. In response, we propose {\it Adaptively Perturbed MD} (APMD), which adjusts the magnitude of the perturbation by repeatedly updating the slingshot strategy at a predefined interval. This innovation empowers us to find a Nash equilibrium of the underlying game with guaranteed rates. Empirical demonstrations affirm that our algorithm exhibits significantly accelerated convergence.
Authors: Shachi Deshpande, Volodymyr Kuleshov
Abstract: Propensity scores are commonly used to estimate treatment effects from observational data. We argue that the probabilistic output of a learned propensity score model should be calibrated -- i.e., a predictive treatment probability of 90% should correspond to 90% of individuals being assigned the treatment group -- and we propose simple recalibration techniques to ensure this property. We prove that calibration is a necessary condition for unbiased treatment effect estimation when using popular inverse propensity weighted and doubly robust estimators. We derive error bounds on causal effect estimates that directly relate to the quality of uncertainties provided by the probabilistic propensity score model and show that calibration strictly improves this error bound while also avoiding extreme propensity weights. We demonstrate improved causal effect estimation with calibrated propensity scores in several tasks including high-dimensional image covariates and genome-wide association studies (GWASs). Calibrated propensity scores improve the speed of GWAS analysis by more than two-fold by enabling the use of simpler models that are faster to train.
Authors: Sehoon Kim, Coleman Hooper, Amir Gholami, Zhen Dong, Xiuyu Li, Sheng Shen, Michael W. Mahoney, Kurt Keutzer
Abstract: Generative Large Language Models (LLMs) have demonstrated remarkable results for a wide range of tasks. However, deploying these models for inference has been a significant challenge due to their unprecedented resource requirements. This has forced existing deployment frameworks to use multi-GPU inference pipelines, which are often complex and costly, or to use smaller and less performant models. In this work, we demonstrate that the main bottleneck for generative inference with LLMs is memory bandwidth, rather than compute, specifically for single batch inference. While quantization has emerged as a promising solution by representing weights with reduced precision, previous efforts have often resulted in notable performance degradation. To address this, we introduce SqueezeLLM, a post-training quantization framework that not only enables lossless compression to ultra-low precisions of up to 3-bit, but also achieves higher quantization performance under the same memory constraint. Our framework incorporates two novel ideas: (i) sensitivity-based non-uniform quantization, which searches for the optimal bit precision assignment based on second-order information; and (ii) the Dense-and-Sparse decomposition that stores outliers and sensitive weight values in an efficient sparse format. When applied to the LLaMA models, our 3-bit quantization significantly reduces the perplexity gap from the FP16 baseline by up to 2.1x as compared to the state-of-the-art methods with the same memory requirement. Furthermore, when deployed on an A6000 GPU, our quantized models achieve up to 2.3x speedup compared to the baseline. Our code is available at https://github.com/SqueezeAILab/SqueezeLLM.
Authors: Xiaoyu Zhang, Saurabh Gupta
Abstract: Autonomous agriculture applications (e.g., inspection, phenotyping, plucking fruits) require manipulating the plant foliage to look behind the leaves and the branches. Partial visibility, extreme clutter, thin structures, and unknown geometry and dynamics for plants make such manipulation challenging. We tackle these challenges through data-driven methods. We use self-supervision to train SRPNet, a neural network that predicts what space is revealed on execution of a candidate action on a given plant. We use SRPNet with the cross-entropy method to predict actions that are effective at revealing space beneath plant foliage. Furthermore, as SRPNet does not just predict how much space is revealed but also where it is revealed, we can execute a sequence of actions that incrementally reveal more and more space beneath the plant foliage. We experiment with a synthetic (vines) and a real plant (Dracaena) on a physical test-bed across 5 settings including 2 settings that test generalization to novel plant configurations. Our experiments reveal the effectiveness of our overall method, PPG, over a competitive hand-crafted exploration method, and the effectiveness of SRPNet over a hand-crafted dynamics model and relevant ablations.
Authors: Giovanni Catania, Aur\'elien Decelle, Beatriz Seoane
Abstract: We characterize the equilibrium properties of a model of $y$ coupled binary perceptrons in the teacher-student scenario, subject to a suitable cost function, with an explicit ferromagnetic coupling proportional to the Hamming distance between the students' weights. In contrast to recent works, we analyze a more general setting in which thermal noise is present that affects each student's generalization performance. In the nonzero temperature regime, we find that the coupling of replicas leads to a bend of the phase diagram towards smaller values of $\alpha$: This suggests that the free entropy landscape gets smoother around the solution with perfect generalization (i.e., the teacher) at a fixed fraction of examples, allowing standard thermal updating algorithms such as Simulated Annealing to easily reach the teacher solution and avoid getting trapped in metastable states as it happens in the unreplicated case, even in the computationally \textit{easy} regime of the inference phase diagram. These results provide additional analytic and numerical evidence for the recently conjectured Bayes-optimal property of Replicated Simulated Annealing (RSA) for a sufficient number of replicas. From a learning perspective, these results also suggest that multiple students working together (in this case reviewing the same data) are able to learn the same rule both significantly faster and with fewer examples, a property that could be exploited in the context of cooperative and federated learning.
Authors: Heinrich Peters, Sandra Matz
Abstract: Large Language Models (LLMs) demonstrate increasingly human-like abilities across a wide variety of tasks. In this paper, we investigate whether LLMs like ChatGPT can accurately infer the psychological dispositions of social media users and whether their ability to do so varies across socio-demographic groups. Specifically, we test whether GPT-3.5 and GPT-4 can derive the Big Five personality traits from users' Facebook status updates in a zero-shot learning scenario. Our results show an average correlation of r = .29 (range = [.22, .33]) between LLM-inferred and self-reported trait scores - a level of accuracy that is similar to that of supervised machine learning models specifically trained to infer personality. Our findings also highlight heterogeneity in the accuracy of personality inferences across different age groups and gender categories: predictions were found to be more accurate for women and younger individuals on several traits, suggesting a potential bias stemming from the underlying training data or differences in online self-expression. The ability of LLMs to infer psychological dispositions from user-generated text has the potential to democratize access to cheap and scalable psychometric assessments for both researchers and practitioners. On the one hand, this democratization might facilitate large-scale research of high ecological validity and spark innovation in personalized services. On the other hand, it also raises ethical concerns regarding user privacy and self-determination, highlighting the need for stringent ethical frameworks and regulation.
Authors: Ralph Peeters, Christian Bizer
Abstract: Entity Matching is the task of deciding whether two entity descriptions refer to the same real-world entity and is a central step in most data integration pipelines. Many state-of-the-art entity matching methods rely on pre-trained language models (PLMs) such as BERT or RoBERTa. Two major drawbacks of these models for entity matching are that (i) the models require significant amounts of task-specific training data and (ii) the fine-tuned models are not robust concerning out-of-distribution entities. This paper investigates using generative large language models (LLMs) as a less task-specific training data-dependent and more robust alternative to PLM-based matchers. Our study covers hosted and open-source LLMs, which can be run locally. We evaluate these models in a zero-shot scenario and a scenario where task-specific training data is available. We compare different prompt designs and the prompt sensitivity of the models and show that there is no single best prompt but needs to be tuned for each model/dataset combination. We further investigate (i) the selection of in-context demonstrations, (ii) the generation of matching rules, as well as (iii) fine-tuning a hosted LLM using the same pool of training data. Our experiments show that the best LLMs require no or only a few training examples to perform similarly to PLMs that were fine-tuned using thousands of examples. LLM-based matchers further exhibit higher robustness to unseen entities. We show that GPT4 can generate structured explanations for matching decisions. The model can automatically identify potential causes of matching errors by analyzing explanations of wrong decisions. We demonstrate that the model can generate meaningful textual descriptions of the identified error classes, which can help data engineers improve entity matching pipelines.
Authors: Wujiang Xu, Xuying Ning, Wenfang Lin, Mingming Ha, Qiongxu Ma, Qianqiao Liang, Xuewen Tao, Linxun Chen, Bing Han, Minnan Luo
Abstract: Cross-domain sequential recommendation (CDSR) aims to address the data sparsity problems that exist in traditional sequential recommendation (SR) systems. The existing approaches aim to design a specific cross-domain unit that can transfer and propagate information across multiple domains by relying on overlapping users with abundant behaviors. However, in real-world recommender systems, CDSR scenarios usually consist of a majority of long-tailed users with sparse behaviors and cold-start users who only exist in one domain. This leads to a drop in the performance of existing CDSR methods in the real-world industry platform. Therefore, improving the consistency and effectiveness of models in open-world CDSR scenarios is crucial for constructing CDSR models (\textit{1st} CH). Recently, some SR approaches have utilized auxiliary behaviors to complement the information for long-tailed users. However, these multi-behavior SR methods cannot deliver promising performance in CDSR, as they overlook the semantic gap between target and auxiliary behaviors, as well as user interest deviation across domains (\textit{2nd} CH).
Authors: Tom S\"uhr, Florian E. Dorner, Samira Samadi, Augustin Kelava
Abstract: With large language models (LLMs) like GPT-4 appearing to behave increasingly human-like in text-based interactions, it has become popular to attempt to evaluate personality traits of LLMs using questionnaires originally developed for humans. While reusing measures is a resource-efficient way to evaluate LLMs, careful adaptations are usually required to ensure that assessment results are valid even across human subpopulations. In this work, we provide evidence that LLMs' responses to personality tests systematically deviate from human responses, implying that the results of these tests cannot be interpreted in the same way. Concretely, reverse-coded items ("I am introverted" vs. "I am extraverted") are often both answered affirmatively. Furthermore, variation across prompts designed to "steer" LLMs to simulate particular personality types does not follow the clear separation into five independent personality factors from human samples. In light of these results, we believe that it is important to investigate tests' validity for LLMs before drawing strong conclusions about potentially ill-defined concepts like LLMs' "personality".
Authors: Nakyeong Yang, Taegwan Kang, Jungkyu Choi, Honglak Lee, Kyomin Jung
Abstract: Instruction-following language models often show undesirable biases. These undesirable biases may be accelerated in the real-world usage of language models, where a wide range of instructions is used through zero-shot example prompting. To solve this problem, we first define the bias neuron, which significantly affects biased outputs, and prove its existence empirically. Furthermore, we propose a novel and practical bias mitigation method, CRISPR, to eliminate bias neurons of language models in instruction-following settings. CRISPR automatically determines biased outputs and categorizes neurons that affect the biased outputs as bias neurons using an explainability method. Experimental results demonstrate the effectiveness of our method in mitigating biases under zero-shot instruction-following settings without losing the model's task performance and existing knowledge. The experimental results reveal the generalizability of our method as it shows robustness under various instructions and datasets. Surprisingly, our method can mitigate the bias in language models by eliminating only a few neurons (at least three).
Authors: Omri Avrahami, Amir Hertz, Yael Vinker, Moab Arar, Shlomi Fruchter, Ohad Fried, Daniel Cohen-Or, Dani Lischinski
Abstract: Recent advances in text-to-image generation models have unlocked vast potential for visual creativity. However, the users that use these models struggle with the generation of consistent characters, a crucial aspect for numerous real-world applications such as story visualization, game development, asset design, advertising, and more. Current methods typically rely on multiple pre-existing images of the target character or involve labor-intensive manual processes. In this work, we propose a fully automated solution for consistent character generation, with the sole input being a text prompt. We introduce an iterative procedure that, at each stage, identifies a coherent set of images sharing a similar identity and extracts a more consistent identity from this set. Our quantitative analysis demonstrates that our method strikes a better balance between prompt alignment and identity consistency compared to the baseline methods, and these findings are reinforced by a user study. To conclude, we showcase several practical applications of our approach.
Authors: Vanessa Graber, Michele Ronchi, Celsa Pardo-Araujo, Nanda Rea
Abstract: We combine pulsar population synthesis with simulation-based inference (SBI) to constrain the magnetorotational properties of isolated Galactic radio pulsars. We first develop a framework to model neutron star birth properties and their dynamical and magnetorotational evolution. We specifically sample initial magnetic field strengths, $B$, and spin periods, $P$, from lognormal distributions and capture the late-time magnetic field decay with a power law. Each lognormal is described by a mean, $\mu_{\log B}, \mu_{\log P}$, and standard deviation, $\sigma_{\log B}, \sigma_{\log P}$, while the power law is characterized by the index, $a_{\rm late}$. We subsequently model the stars' radio emission and observational biases to mimic detections with three radio surveys, and we produce a large database of synthetic $P$--$\dot{P}$ diagrams by varying our five magnetorotational input parameters. We then follow an SBI approach that focuses on neural posterior estimation and train deep neural networks to infer the parameters' posterior distributions. After successfully validating these individual neural density estimators on simulated data, we use an ensemble of networks to infer the posterior distributions for the observed pulsar population. We obtain $\mu_{\log B} = 13.10^{+0.08}_{-0.10}$, $\sigma_{\log B} = 0.45^{+0.05}_{-0.05}$ and $\mu_{\log P} = -1.00^{+0.26}_{-0.21}$, $\sigma_{\log P} = 0.38^{+0.33}_{-0.18}$ for the lognormal distributions and $a_{\rm late} = -1.80^{+0.65}_{-0.61}$ for the power law at the $95\%$ credible interval. We contrast our results with previous studies and highlight uncertainties of the inferred $a_{\rm late}$ value. Our approach represents a crucial step toward robust statistical inference for complex population synthesis frameworks and forms the basis for future multiwavelength analyses of Galactic pulsars.
Authors: Chen Qian, Yufan Dang, Jiahao Li, Wei Liu, Zihao Xie, Yifei Wang, Weize Chen, Cheng Yang, Xin Cong, Xiaoyin Che, Zhiyuan Liu, Maosong Sun
Abstract: Recent advancements in large language models (LLMs) have brought significant changes to various domains, especially through LLM-driven autonomous agents. A representative scenario is in software development, where LLM agents demonstrate efficient collaboration, task division, and assurance of software quality, markedly reducing the need for manual involvement. However, these agents frequently perform a variety of tasks independently, without benefiting from past experiences, which leads to repeated mistakes and inefficient attempts in multi-step task execution. To this end, we introduce Experiential Co-Learning, a novel LLM-agent learning framework in which instructor and assistant agents gather shortcut-oriented experiences from their historical trajectories and use these past experiences for future task execution. The extensive experiments demonstrate that the framework enables agents to tackle unseen software-developing tasks more effectively. We anticipate that our insights will guide LLM agents towards enhanced autonomy and contribute to their evolutionary growth in cooperative learning. The code and data are available at https://github.com/OpenBMB/ChatDev.
Authors: Hengrui Cai, Shengjie Liu, Rui Song
Abstract: This paper explores the causal reasoning of large language models (LLMs) to enhance their interpretability and reliability in advancing artificial intelligence. Despite the proficiency of LLMs in a range of tasks, their potential for understanding causality requires further exploration. We propose a novel causal attribution model that utilizes ``do-operators" for constructing counterfactual scenarios, allowing us to systematically quantify the influence of input numerical data and LLMs' pre-existing knowledge on their causal reasoning processes. Our newly developed experimental setup assesses LLMs' reliance on contextual information and inherent knowledge across various domains. Our evaluation reveals that LLMs' causal reasoning ability mainly depends on the context and domain-specific knowledge provided. In the absence of such knowledge, LLMs can still maintain a degree of causal reasoning using the available numerical data, albeit with limitations in the calculations. This motivates the proposed fine-tuned LLM for pairwise causal discovery, effectively leveraging both knowledge and numerical information.
Authors: Ying Sheng, Shiyi Cao, Dacheng Li, Banghua Zhu, Zhuohan Li, Danyang Zhuo, Joseph E. Gonzalez, Ion Stoica
Abstract: High-demand LLM inference services (e.g., ChatGPT and BARD) support a wide range of requests from short chat conversations to long document reading. To ensure that all client requests are processed fairly, most major LLM inference services have request rate limits, to ensure that no client can dominate the request queue. However, this rudimentary notion of fairness also results in under-utilization of the resources and poor client experience when there is spare capacity. While there is a rich literature on fair scheduling, serving LLMs presents new challenges due to their unpredictable request lengths and their unique batching characteristics on parallel accelerators. This paper introduces the definition of LLM serving fairness based on a cost function that accounts for the number of input and output tokens processed. To achieve fairness in serving, we propose a novel scheduling algorithm, the Virtual Token Counter (VTC), a fair scheduler based on the continuous batching mechanism. We prove a 2x tight upper bound on the service difference between two backlogged clients, adhering to the requirement of work-conserving. Through extensive experiments, we demonstrate the superior performance of VTC in ensuring fairness, especially in contrast to other baseline methods, which exhibit shortcomings under various conditions. The reproducible code is available at https://github.com/Ying1123/VTC-artifact
Authors: Dyah A. M. G. Wisnu, Stefano Rini, Ryandhimas E. Zezario, Hsin-Min Wang, Yu Tsao
Abstract: This paper introduces HAAQI-Net, a non-intrusive deep learning model for music audio quality assessment tailored for hearing aid users. Unlike traditional methods like the Hearing Aid Audio Quality Index (HAAQI), which rely on intrusive comparisons to a reference signal, HAAQI-Net offers a more accessible and efficient alternative. Using a bidirectional Long Short-Term Memory (BLSTM) architecture with attention mechanisms and features from the pre-trained BEATs model, HAAQI-Net predicts HAAQI scores directly from music audio clips and hearing loss patterns. Results show HAAQI-Net's effectiveness, with predicted scores achieving a Linear Correlation Coefficient (LCC) of 0.9368, a Spearman's Rank Correlation Coefficient (SRCC) of 0.9486, and a Mean Squared Error (MSE) of 0.0064, reducing inference time from 62.52 seconds to 2.54 seconds. Although effective, feature extraction via the large BEATs model incurs computational overhead. To address this, a knowledge distillation strategy creates a student distillBEATs model, distilling information from the teacher BEATs model during HAAQI-Net training, reducing required parameters. The distilled HAAQI-Net maintains strong performance with an LCC of 0.9071, an SRCC of 0.9307, and an MSE of 0.0091, while reducing parameters by 75.85% and inference time by 96.46%. This reduction enhances HAAQI-Net's efficiency and scalability, making it viable for real-world music audio quality assessment in hearing aid settings. This work also opens avenues for further research into optimizing deep learning models for specific applications, contributing to audio signal processing and quality assessment by providing insights into developing efficient and accurate models for practical applications in hearing aid technology.
Authors: Aimee Maurais, Youssef Marzouk
Abstract: We introduce a new mean-field ODE and corresponding interacting particle systems (IPS) for sampling from an unnormalized target density. The IPS are gradient-free, available in closed form, and only require the ability to sample from a reference density and compute the (unnormalized) target-to-reference density ratio. The mean-field ODE is obtained by solving a Poisson equation for a velocity field that transports samples along the geometric mixture of the two densities, which is the path of a particular Fisher-Rao gradient flow. We employ a RKHS ansatz for the velocity field, which makes the Poisson equation tractable and enables discretization of the resulting mean-field ODE over finite samples. The mean-field ODE can be additionally be derived from a discrete-time perspective as the limit of successive linearizations of the Monge-Amp\`ere equations within a framework known as sample-driven optimal transport. We introduce a stochastic variant of our approach and demonstrate empirically that our IPS can produce high-quality samples from varied target distributions, outperforming comparable gradient-free particle systems and competitive with gradient-based alternatives.
Authors: Peter Hase, Mohit Bansal, Peter Clark, Sarah Wiegreffe
Abstract: How can we train models to perform well on hard test data when hard training data is by definition difficult to label correctly? This question has been termed the scalable oversight problem and has drawn increasing attention as language models have continually improved. In this paper, we present the surprising conclusion that current pretrained language models often generalize relatively well from easy to hard data, even performing as well as oracle models finetuned on hard data. We demonstrate this kind of easy-to-hard generalization using simple finetuning methods like in-context learning, linear classifier heads, and QLoRA for seven different measures of datapoint hardness, including six empirically diverse human hardness measures (like grade level) and one model-based measure (loss-based). Furthermore, we show that even if one cares most about model performance on hard data, it can be better to collect easy data rather than hard data for finetuning, since hard data is generally noisier and costlier to collect. Our experiments use open models up to 70b in size and four publicly available question-answering datasets with questions ranging in difficulty from 3rd grade science questions to college level STEM questions and general-knowledge trivia. We conclude that easy-to-hard generalization in LMs is surprisingly strong for the tasks studied. Our code is available at: https://github.com/allenai/easy-to-hard-generalization
URLs: https://github.com/allenai/easy-to-hard-generalization
Authors: Tom Dooney, Lyana Curier, Daniel Tan, Melissa Lopez, Chris Van Den Broeck, Stefano Bromuri
Abstract: Simulating realistic time-domain observations of gravitational waves (GWs) and GW detector glitches can help in advancing GW data analysis. Simulated data can be used in downstream tasks by augmenting datasets for signal searches, balancing data sets for machine learning, and validating detection schemes. In this work, we present Conditional Derivative GAN (cDVGAN), a novel conditional model in the Generative Adversarial Network framework for simulating multiple classes of time-domain observations that represent gravitational waves (GWs) and detector glitches. cDVGAN can also generate generalized hybrid samples that span the variation between classes through interpolation in the conditioned class vector. cDVGAN introduces an additional player into the typical 2-player adversarial game of GANs, where an auxiliary discriminator analyzes the first-order derivative time-series. Our results show that this provides synthetic data that better captures the features of the original data. cDVGAN conditions on three classes, two denoised from LIGO blip and tomte glitch events from its 3rd observing run (O3), and the third representing binary black hole (BBH) mergers. Our proposed cDVGAN outperforms 4 different baseline GAN models in replicating the features of the three classes. Specifically, our experiments show that training convolutional neural networks (CNNs) with our cDVGAN-generated data improves the detection of samples embedded in detector noise beyond the synthetic data from other state-of-the-art GAN models. Our best synthetic dataset yields as much as a 4.2% increase in area-under-the-curve (AUC) performance compared to synthetic datasets from baseline GANs. Moreover, training the CNN with hybrid samples from our cDVGAN outperforms CNNs trained only on the standard classes, when identifying real samples embedded in LIGO detector background (4% AUC improvement for cDVGAN).
Authors: Xiangru Tang, Qiao Jin, Kunlun Zhu, Tongxin Yuan, Yichi Zhang, Wangchunshu Zhou, Meng Qu, Yilun Zhao, Jian Tang, Zhuosheng Zhang, Arman Cohan, Zhiyong Lu, Mark Gerstein
Abstract: Intelligent agents powered by large language models (LLMs) have demonstrated substantial promise in autonomously conducting experiments and facilitating scientific discoveries across various disciplines. While their capabilities are promising, these agents, called scientific LLM agents, also introduce novel vulnerabilities that demand careful consideration for safety. However, there exists a notable gap in the literature, as there has been no comprehensive exploration of these vulnerabilities. This perspective paper fills this gap by conducting a thorough examination of vulnerabilities in LLM-based agents within scientific domains, shedding light on potential risks associated with their misuse and emphasizing the need for safety measures. We begin by providing a comprehensive overview of the potential risks inherent to scientific LLM agents, taking into account user intent, the specific scientific domain, and their potential impact on the external environment. Then, we delve into the origins of these vulnerabilities and provide a scoping review of the limited existing works. Based on our analysis, we propose a triadic framework involving human regulation, agent alignment, and an understanding of environmental feedback (agent regulation) to mitigate these identified risks. Furthermore, we highlight the limitations and challenges associated with safeguarding scientific agents and advocate for the development of improved models, robust benchmarks, and comprehensive regulations to address these issues effectively.
Authors: Michael E. Sander, Raja Giryes, Taiji Suzuki, Mathieu Blondel, Gabriel Peyr\'e
Abstract: Transformers have achieved state-of-the-art performance in language modeling tasks. However, the reasons behind their tremendous success are still unclear. In this paper, towards a better understanding, we train a Transformer model on a simple next token prediction task, where sequences are generated as a first-order autoregressive process $s_{t+1} = W s_t$. We show how a trained Transformer predicts the next token by first learning $W$ in-context, then applying a prediction mapping. We call the resulting procedure in-context autoregressive learning. More precisely, focusing on commuting orthogonal matrices $W$, we first show that a trained one-layer linear Transformer implements one step of gradient descent for the minimization of an inner objective function, when considering augmented tokens. When the tokens are not augmented, we characterize the global minima of a one-layer diagonal linear multi-head Transformer. Importantly, we exhibit orthogonality between heads and show that positional encoding captures trigonometric relations in the data. On the experimental side, we consider the general case of non-commuting orthogonal matrices and generalize our theoretical findings.
Authors: Shengxiang Hu, Guobing Zou, Song Yang, Yanglan Gan, Bofeng Zhang, Yixin Chen
Abstract: In service-oriented architectures, accurately predicting the Quality of Service (QoS) is crucial for maintaining reliability and enhancing user satisfaction. However, significant challenges remain due to existing methods always overlooking high-order latent collaborative relationships between users and services and failing to dynamically adjust feature learning for every specific user-service invocation, which are critical for learning accurate features. Additionally, reliance on RNNs for capturing QoS evolution hampers models' ability to detect long-term trends due to difficulties in managing long-range dependencies. To address these challenges, we propose the \underline{T}arget-Prompt \underline{O}nline \underline{G}raph \underline{C}ollaborative \underline{L}earning (TOGCL) framework for temporal-aware QoS prediction. TOGCL leverages a dynamic user-service invocation graph to model historical interactions, providing a comprehensive representation of user-service relationships. Building on this graph, it develops a target-prompt graph attention network to extract online deep latent features of users and services at each time slice, simultaneously considering implicit collaborative relationships between target users/services and their neighbors, as well as relevant historical QoS values. Additionally, a multi-layer Transformer encoder is employed to uncover temporal feature evolution patterns of users and services, leading to temporal-aware QoS prediction. Extensive experiments conducted on the WS-DREAM dataset demonstrate that our proposed TOGCL framework significantly outperforms state-of-the-art methods across multiple metrics, achieving improvements of up to 38.80\%. These results underscore the effectiveness of the TOGCL framework for precise temporal QoS prediction.
Authors: Shir Ashury-Tahan, Ariel Gera, Benjamin Sznajder, Leshem Choshen, Liat Ein-Dor, Eyal Shnarch
Abstract: Model selection for a given target task can be costly, as it may entail extensive annotation of the quality of outputs of different models. We introduce DiffUse, an efficient method to make an informed decision between candidate text generation models based on preference annotations. DiffUse reduces the required amount of annotations, thus saving valuable time and resources in performing evaluation. DiffUse intelligently selects instances by clustering embeddings that represent the semantic differences between model outputs. Thus, it is able to identify a subset of examples that are more informative for preference decisions. Our method is model-agnostic, and can be applied to any text generation model for selecting between models, prompts and configurations. Moreover, we propose a practical iterative approach for dynamically determining how many instances to annotate. In a series of experiments over hundreds of model pairs, we demonstrate that DiffUse can dramatically reduce the required number of annotations -- by up to 75% -- while maintaining high evaluation reliability.
Authors: S Ashwin Hebbar, Sravan Kumar Ankireddy, Hyeji Kim, Sewoong Oh, Pramod Viswanath
Abstract: Progress in designing channel codes has been driven by human ingenuity and, fittingly, has been sporadic. Polar codes, developed on the foundation of Arikan's polarization kernel, represent the latest breakthrough in coding theory and have emerged as the state-of-the-art error-correction code for short-to-medium block length regimes. In an effort to automate the invention of good channel codes, especially in this regime, we explore a novel, non-linear generalization of Polar codes, which we call DeepPolar codes. DeepPolar codes extend the conventional Polar coding framework by utilizing a larger kernel size and parameterizing these kernels and matched decoders through neural networks. Our results demonstrate that these data-driven codes effectively leverage the benefits of a larger kernel size, resulting in enhanced reliability when compared to both existing neural codes and conventional Polar codes.
Authors: Yaoyiran Li, Anna Korhonen, Ivan Vuli\'c
Abstract: Recent work has shown that, while large language models (LLMs) demonstrate strong word translation or bilingual lexicon induction (BLI) capabilities in few-shot setups, they still cannot match the performance of 'traditional' mapping-based approaches in the unsupervised scenario where no seed translation pairs are available, especially for lower-resource languages. To address this challenge with LLMs, we propose self-augmented in-context learning (SAIL) for unsupervised BLI: starting from a zero-shot prompt, SAIL iteratively induces a set of high-confidence word translation pairs for in-context learning (ICL) from an LLM, which it then reapplies to the same LLM in the ICL fashion. Our method shows substantial gains over zero-shot prompting of LLMs on two established BLI benchmarks spanning a wide range of language pairs, also outperforming mapping-based baselines across the board. In addition to achieving state-of-the-art unsupervised BLI performance, we also conduct comprehensive analyses on SAIL and discuss its limitations.
Authors: Pascal Pernot
Abstract: Average calibration of the (variance-based) prediction uncertainties of machine learning regression tasks can be tested in two ways: one is to estimate the calibration error (CE) as the difference between the mean absolute error (MSE) and the mean variance (MV); the alternative is to compare the mean squared z-scores (ZMS) to 1. The problem is that both approaches might lead to different conclusions, as illustrated in this study for an ensemble of datasets from the recent machine learning uncertainty quantification (ML-UQ) literature. It is shown that the estimation of MV, MSE and their confidence intervals becomes unreliable for heavy-tailed uncertainty and error distributions, which seems to be a frequent feature of ML-UQ datasets. By contrast, the ZMS statistic is less sensitive and offers the most reliable approach in this context. Unfortunately, the same problem is expected to affect also conditional calibrations statistics, such as the popular ENCE, and very likely post-hoc calibration methods based on similar statistics. Several solutions to circumvent the outlined problems are proposed.
Authors: Amit Dhurandhar, Rahul Nair, Moninder Singh, Elizabeth Daly, Karthikeyan Natesan Ramamurthy
Abstract: Evaluation and ranking of large language models (LLMs) has become an important problem with the proliferation of these models and their impact. Evaluation methods either require human responses which are expensive to acquire or use pairs of LLMs to evaluate each other which can be unreliable. In this paper, we provide a novel perspective where, given a dataset of prompts (viz. questions, instructions, etc.) and a set of LLMs, we rank them without access to any ground truth or reference responses. Inspired by real life where both an expert and a knowledgeable person can identify a novice our main idea is to consider triplets of models, where each one of them evaluates the other two, correctly identifying the worst model in the triplet with high probability. We also analyze our idea and provide sufficient conditions for it to succeed. Applying this idea repeatedly, we propose two methods to rank LLMs. In experiments on different generative tasks (summarization, multiple-choice, and dialog), our methods reliably recover close to true rankings without reference data. This points to a viable low-resource mechanism for practical use.
Authors: Shen Li, Liuyi Yao, Jinyang Gao, Lan Zhang, Yaliang Li
Abstract: To support various applications, a prevalent and efficient approach for business owners is leveraging their valuable datasets to fine-tune a pre-trained LLM through the API provided by LLM owners or cloud servers. However, this process carries a substantial risk of model misuse, potentially resulting in severe economic consequences for business owners. Thus, safeguarding the copyright of these customized models during LLM fine-tuning has become an urgent practical requirement, but there are limited existing solutions to provide such protection. To tackle this pressing issue, we propose a novel watermarking approach named ``Double-I watermark''. Specifically, based on the instruct-tuning data, two types of backdoor data paradigms are introduced with trigger in the instruction and the input, respectively. By leveraging LLM's learning capability to incorporate customized backdoor samples into the dataset, the proposed approach effectively injects specific watermarking information into the customized model during fine-tuning, which makes it easy to inject and verify watermarks in commercial scenarios. We evaluate the proposed "Double-I watermark" under various fine-tuning methods, demonstrating its harmlessness, robustness, uniqueness, imperceptibility, and validity through both quantitative and qualitative analyses.
Authors: Nitesh Kumar, Usashi Chatterjee, Steven Schockaert
Abstract: Conceptual spaces represent entities in terms of their primitive semantic features. Such representations are highly valuable but they are notoriously difficult to learn, especially when it comes to modelling perceptual and subjective features. Distilling conceptual spaces from Large Language Models (LLMs) has recently emerged as a promising strategy, but existing work has been limited to probing pre-trained LLMs using relatively simple zero-shot strategies. We focus in particular on the task of ranking entities according to a given conceptual space dimension. Unfortunately, we cannot directly fine-tune LLMs on this task, because ground truth rankings for conceptual space dimensions are rare. We therefore use more readily available features as training data and analyse whether the ranking capabilities of the resulting models transfer to perceptual and subjective features. We find that this is indeed the case, to some extent, but having at least some perceptual and subjective features in the training data seems essential for achieving the best results.
Authors: Xiaoyu Zhang, Matthew Chang, Pranav Kumar, Saurabh Gupta
Abstract: A common failure mode for policies trained with imitation is compounding execution errors at test time. When the learned policy encounters states that are not present in the expert demonstrations, the policy fails, leading to degenerate behavior. The Dataset Aggregation, or DAgger approach to this problem simply collects more data to cover these failure states. However, in practice, this is often prohibitively expensive. In this work, we propose Diffusion Meets DAgger (DMD), a method to reap the benefits of DAgger without the cost for eye-in-hand imitation learning problems. Instead of collecting new samples to cover out-of-distribution states, DMD uses recent advances in diffusion models to synthesize these samples. This leads to robust performance from few demonstrations. We compare DMD against behavior cloning baseline across four tasks: pushing, stacking, pouring, and shirt hanging. In pushing, DMD achieves 80% success rate with as few as 8 expert demonstrations, where naive behavior cloning reaches only 20%. In stacking, DMD succeeds on average 92% of the time across 5 cups, versus 40% for BC. When pouring coffee beans, DMD transfers to another cup successfully 80% of the time. Finally, DMD attains 90% success rate for hanging shirt on a clothing rack.
Authors: Shaolei Zhang, Tian Yu, Yang Feng
Abstract: Large Language Models (LLMs) sometimes suffer from producing hallucinations, especially LLMs may generate untruthful responses despite knowing the correct knowledge. Activating the truthfulness within LLM is the key to fully unlocking LLM's knowledge potential. In this paper, we propose TruthX, an inference-time intervention method to activate the truthfulness of LLM by identifying and editing the features within LLM's internal representations that govern the truthfulness. TruthX employs an auto-encoder to map LLM's representations into semantic and truthful latent spaces respectively, and applies contrastive learning to identify a truthful editing direction within the truthful space. During inference, by editing LLM's internal representations in truthful space, TruthX effectively enhances the truthfulness of LLM. Experiments show that TruthX improves the truthfulness of 13 advanced LLMs by an average of 20% on TruthfulQA benchmark. Further analyses suggest that TruthX can control LLM to produce truthful or hallucinatory responses via editing only one vector in LLM's internal representations.
Authors: Adithya Bhaskar, Dan Friedman, Danqi Chen
Abstract: Prior work has found that pretrained language models (LMs) fine-tuned with different random seeds can achieve similar in-domain performance but generalize differently on tests of syntactic generalization. In this work, we show that, even within a single model, we can find multiple subnetworks that perform similarly in-domain, but generalize vastly differently. To better understand these phenomena, we investigate if they can be understood in terms of "competing subnetworks": the model initially represents a variety of distinct algorithms, corresponding to different subnetworks, and generalization occurs when it ultimately converges to one. This explanation has been used to account for generalization in simple algorithmic tasks ("grokking"). Instead of finding competing subnetworks, we find that all subnetworks -- whether they generalize or not -- share a set of attention heads, which we refer to as the heuristic core. Further analysis suggests that these attention heads emerge early in training and compute shallow, non-generalizing features. The model learns to generalize by incorporating additional attention heads, which depend on the outputs of the "heuristic" heads to compute higher-level features. Overall, our results offer a more detailed picture of the mechanisms for syntactic generalization in pretrained LMs.
Authors: Ziyu Zhang, Johann Laconte, Daniil Lisus, Timothy D. Barfoot
Abstract: This paper presents a novel method to assess the resilience of the Iterative Closest Point (ICP) algorithm via deep-learning-based attacks on lidar point clouds. For safety-critical applications such as autonomous navigation, ensuring the resilience of algorithms prior to deployments is of utmost importance. The ICP algorithm has become the standard for lidar-based localization. However, the pose estimate it produces can be greatly affected by corruption in the measurements. Corruption can arise from a variety of scenarios such as occlusions, adverse weather, or mechanical issues in the sensor. Unfortunately, the complex and iterative nature of ICP makes assessing its resilience to corruption challenging. While there have been efforts to create challenging datasets and develop simulations to evaluate the resilience of ICP empirically, our method focuses on finding the maximum possible ICP pose error using perturbation-based adversarial attacks. The proposed attack induces significant pose errors on ICP and outperforms baselines more than 88% of the time across a wide range of scenarios. As an example application, we demonstrate that our attack can be used to identify areas on a map where ICP is particularly vulnerable to corruption in the measurements.
Authors: Vaclav Voracek, Tomas Werner
Abstract: A popular approach to the MAP inference problem in graphical models is to minimize an upper bound obtained from a dual linear programming or Lagrangian relaxation by (block-)coordinate descent. This is also known as convex/convergent message passing; examples are max-sum diffusion and sequential tree-reweighted message passing (TRW-S). Convergence properties of these methods are currently not fully understood. They have been proved to converge to the set characterized by local consistency of active constraints, with unknown convergence rate; however, it was not clear if the iterates converge at all (to any point). We prove a stronger result (conjectured before but never proved): the iterates converge to a fixed point of the method. Moreover, we show that the algorithm terminates within $\mathcal{O}(1/\varepsilon)$ iterations. We first prove this for a version of coordinate descent applied to a general piecewise-affine convex objective. Then we show that several convex message passing methods are special cases of this method. Finally, we show that a slightly different version of coordinate descent can cycle.
Authors: Mustafa Ustuner
Abstract: The high-dimensional feature space of the hyperspectral imagery poses major challenges to the processing and analysis of the hyperspectral data sets. In such a case, dimensionality reduction is necessary to decrease the computational complexity. The random projections open up new ways of dimensionality reduction, especially for large data sets. In this paper, the principal component analysis (PCA) and randomized principal component analysis (R-PCA) for the classification of hyperspectral images using support vector machines (SVM) and light gradient boosting machines (LightGBM) have been investigated. In this experimental research, the number of features was reduced to 20 and 30 for classification of two hyperspectral datasets (Indian Pines and Pavia University). The experimental results demonstrated that PCA outperformed R-PCA for SVM for both datasets, but received close accuracy values for LightGBM. The highest classification accuracies were obtained as 0.9925 and 0.9639 by LightGBM with original features for the Pavia University and Indian Pines, respectively.
Authors: Francesco Paissan, Mirco Ravanelli, Cem Subakan
Abstract: Despite the impressive performance of deep learning models across diverse tasks, their complexity poses challenges for interpretation. This challenge is particularly evident for audio signals, where conveying interpretations becomes inherently difficult. To address this issue, we introduce Listenable Maps for Audio Classifiers (L-MAC), a posthoc interpretation method that generates faithful and listenable interpretations. L-MAC utilizes a decoder on top of a pretrained classifier to generate binary masks that highlight relevant portions of the input audio. We train the decoder with a loss function that maximizes the confidence of the classifier decision on the masked-in portion of the audio while minimizing the probability of model output for the masked-out portion. Quantitative evaluations on both in-domain and out-of-domain data demonstrate that L-MAC consistently produces more faithful interpretations than several gradient and masking-based methodologies. Furthermore, a user study confirms that, on average, users prefer the interpretations generated by the proposed technique.
Authors: Azmine Toushik Wasi, Taki Hasan Rafi, Raima Islam, Dong-Kyu Chae
Abstract: Knowledge Graphs (KGs) have proven essential in information processing and reasoning applications because they link related entities and give context-rich information, supporting efficient information retrieval and knowledge discovery; presenting information flow in a very effective manner. Despite being widely used globally, Bangla is relatively underrepresented in KGs due to a lack of comprehensive datasets, encoders, NER (named entity recognition) models, POS (part-of-speech) taggers, and lemmatizers, hindering efficient information processing and reasoning applications in the language. Addressing the KG scarcity in Bengali, we propose BanglaAutoKG, a pioneering framework that is able to automatically construct Bengali KGs from any Bangla text. We utilize multilingual LLMs to understand various languages and correlate entities and relations universally. By employing a translation dictionary to identify English equivalents and extracting word features from pre-trained BERT models, we construct the foundational KG. To reduce noise and align word embeddings with our goal, we employ graph-based polynomial filters. Lastly, we implement a GNN-based semantic filter, which elevates contextual understanding and trims unnecessary edges, culminating in the formation of the definitive KG. Empirical findings and case studies demonstrate the universal effectiveness of our model, capable of autonomously constructing semantically enriched KGs from any text.
Authors: Zishuo Zhao, Zhixuan Fang, Xuechao Wang, Xi Chen, Yuan Zhou
Abstract: Most concurrent blockchain systems rely heavily on the Proof-of-Work (PoW) or Proof-of-Stake (PoS) mechanisms for decentralized consensus and security assurance. However, the substantial energy expenditure stemming from computationally intensive yet meaningless tasks has raised considerable concerns surrounding traditional PoW approaches, The PoS mechanism, while free of energy consumption, is subject to security and economic issues. Addressing these issues, the paradigm of Proof-of-Useful-Work (PoUW) seeks to employ challenges of practical significance as PoW, thereby imbuing energy consumption with tangible value. While previous efforts in Proof of Learning (PoL) explored the utilization of deep learning model training SGD tasks as PoUW challenges, recent research has revealed its vulnerabilities to adversarial attacks and the theoretical hardness in crafting a byzantine-secure PoL mechanism. In this paper, we introduce the concept of incentive-security that incentivizes rational provers to behave honestly for their best interest, bypassing the existing hardness to design a PoL mechanism with computational efficiency, a provable incentive-security guarantee and controllable difficulty. Particularly, our work is secure against two attacks to the recent work of Jia et al. [2021], and also improves the computational overhead from $\Theta(1)$ to $O(\frac{\log E}{E})$. Furthermore, while most recent research assumes trusted problem providers and verifiers, our design also guarantees frontend incentive-security even when problem providers are untrusted, and verifier incentive-security that bypasses the Verifier's Dilemma. By incorporating ML training into blockchain consensus mechanisms with provable guarantees, our research not only proposes an eco-friendly solution to blockchain systems, but also provides a proposal for a completely decentralized computing power market in the new AI age.
Authors: Taehyeon Kim, Ananda Theertha Suresh, Kishore Papineni, Michael Riley, Sanjiv Kumar, Adrian Benton
Abstract: Despite the remarkable strides made by autoregressive language models, their potential is often hampered by the slow inference speeds inherent in sequential token generation. Blockwise parallel decoding (BPD) was proposed by Stern et al. as a method to improve inference speed of language models by simultaneously predicting multiple future tokens, termed block drafts, which are subsequently verified and conditionally accepted by the autoregressive model. This paper contributes to the understanding and improvement of block drafts in two ways. First, we analyze the token distributions produced by multiple prediction heads. Secondly, we leverage this analysis to develop algorithms to improve BPD inference speed by refining the block drafts using n-gram and neural language models. Experiments demonstrate that refined block drafts yield a +5-21% increase in block efficiency (i.e., the number of accepted tokens from the block draft) across diverse datasets.
Authors: Ali Beikmohammadi, Sarit Khirirat, Sindri Magn\'usson
Abstract: Reinforcement learning has recently gained unprecedented popularity, yet it still grapples with sample inefficiency. Addressing this challenge, federated reinforcement learning (FedRL) has emerged, wherein agents collaboratively learn a single policy by aggregating local estimations. However, this aggregation step incurs significant communication costs. In this paper, we propose CompFedRL, a communication-efficient FedRL approach incorporating both \textit{periodic aggregation} and (direct/error-feedback) compression mechanisms. Specifically, we consider compressed federated $Q$-learning with a generative model setup, where a central server learns an optimal $Q$-function by periodically aggregating compressed $Q$-estimates from local agents. For the first time, we characterize the impact of these two mechanisms (which have remained elusive) by providing a finite-time analysis of our algorithm, demonstrating strong convergence behaviors when utilizing either direct or error-feedback compression. Our bounds indicate improved solution accuracy concerning the number of agents and other federated hyperparameters while simultaneously reducing communication costs. To corroborate our theory, we also conduct in-depth numerical experiments to verify our findings, considering Top-$K$ and Sparsified-$K$ sparsification operators.
Authors: Bryan Lim, Manon Flageat, Antoine Cully
Abstract: Quality-Diversity (QD) approaches are a promising direction to develop open-ended processes as they can discover archives of high-quality solutions across diverse niches. While already successful in many applications, QD approaches usually rely on combining only one or two solutions to generate new candidate solutions. As observed in open-ended processes such as technological evolution, wisely combining large diversity of these solutions could lead to more innovative solutions and potentially boost the productivity of QD search. In this work, we propose to exploit the pattern-matching capabilities of generative models to enable such efficient solution combinations. We introduce In-context QD, a framework of techniques that aim to elicit the in-context capabilities of pre-trained Large Language Models (LLMs) to generate interesting solutions using few-shot and many-shot prompting with quality-diverse examples from the QD archive as context. Applied to a series of common QD domains, In-context QD displays promising results compared to both QD baselines and similar strategies developed for single-objective optimization. Additionally, this result holds across multiple values of parameter sizes and archive population sizes, as well as across domains with distinct characteristics from BBO functions to policy search. Finally, we perform an extensive ablation that highlights the key prompt design considerations that encourage the generation of promising solutions for QD.
Authors: Hanyin Wang, Chufan Gao, Bolun Liu, Qiping Xu, Guleid Hussein, Mohamad El Labban, Kingsley Iheasirim, Hariprasad Korsapati, Chuck Outcalt, Jimeng Sun
Abstract: Proprietary Large Language Models (LLMs) such as GPT-4 and Gemini have demonstrated promising capabilities in clinical text summarization tasks. However, due to patient data privacy concerns and computational costs, many healthcare providers prefer using small, locally-hosted models over external generic LLMs. This study presents a comprehensive domain- and task-specific adaptation process for the open-source LLaMA-2 13 billion parameter model, enabling it to generate high-quality clinical notes from outpatient patient-doctor dialogues. Our process incorporates continued pre-training, supervised fine-tuning, and reinforcement learning from both AI and human feedback. We introduced a new approach, DistillDirect, for performing on-policy reinforcement learning with Gemini 1.0 Pro as the teacher model. Our resulting model, LLaMA-Clinic, can generate clinical notes comparable in quality to those authored by physicians. In a blinded physician reader study, the majority (90.4%) of individual evaluations rated the notes generated by LLaMA-Clinic as "acceptable" or higher across all three criteria: real-world readiness, completeness, and accuracy. In the more challenging "Assessment and Plan" section, LLaMA-Clinic scored higher (4.2/5) in real-world readiness than physician-authored notes (4.1/5). Our cost analysis for inference shows that our LLaMA-Clinic model achieves a 4.375-fold cost reduction compared to an external generic LLM service. Additionally, we highlight key considerations for future clinical note-generation tasks, emphasizing the importance of pre-defining a best-practice note format, rather than relying on LLMs to determine this for clinical practice. We have made our newly created synthetic clinic dialogue-note dataset and the physician feedback dataset publicly available to foster future research.
Authors: Yuyang Zhang, Shahriar Talebi, Na Li
Abstract: In this paper, we focus on learning a linear time-invariant (LTI) model with low-dimensional latent variables but high-dimensional observations. We provide an algorithm that recovers the high-dimensional features, i.e. column space of the observer, embeds the data into low dimensions and learns the low-dimensional model parameters. Our algorithm enjoys a sample complexity guarantee of order $\tilde{\mathcal{O}}(n/\epsilon^2)$, where $n$ is the observation dimension. We further establish a fundamental lower bound indicating this complexity bound is optimal up to logarithmic factors and dimension-independent constants. We show that this inevitable linear factor of $n$ is due to the learning error of the observer's column space in the presence of high-dimensional noises. Extending our results, we consider a meta-learning problem inspired by various real-world applications, where the observer column space can be collectively learned from datasets of multiple LTI systems. An end-to-end algorithm is then proposed, facilitating learning LTI systems from a meta-dataset which breaks the sample complexity lower bound in certain scenarios.
Authors: Florent Bouchard, Ammar Mian, Malik Tiomoko, Guillaume Ginolhac, Fr\'ed\'eric Pascal
Abstract: In this study, we consider the realm of covariance matrices in machine learning, particularly focusing on computing Fr\'echet means on the manifold of symmetric positive definite matrices, commonly referred to as Karcher or geometric means. Such means are leveraged in numerous machine-learning tasks. Relying on advanced statistical tools, we introduce a random matrix theory-based method that estimates Fr\'echet means, which is particularly beneficial when dealing with low sample support and a high number of matrices to average. Our experimental evaluation, involving both synthetic and real-world EEG and hyperspectral datasets, shows that we largely outperform state-of-the-art methods.
Authors: Abbi Abdel-Rehim, Hector Zenil, Oghenejokpeme Orhobor, Marie Fisher, Ross J. Collins, Elizabeth Bourne, Gareth W. Fearnley, Emma Tate, Holly X. Smith, Larisa N. Soldatova, Ross D. King
Abstract: Large language models (LLMs) have transformed AI and achieved breakthrough performance on a wide range of tasks that require human intelligence. In science, perhaps the most interesting application of LLMs is for hypothesis formation. A feature of LLMs, which results from their probabilistic structure, is that the output text is not necessarily a valid inference from the training text. These are 'hallucinations', and are a serious problem in many applications. However, in science, hallucinations may be useful: they are novel hypotheses whose validity may be tested by laboratory experiments. Here we experimentally test the use of LLMs as a source of scientific hypotheses using the domain of breast cancer treatment. We applied the LLM GPT4 to hypothesize novel pairs of FDA-approved non-cancer drugs that target the MCF7 breast cancer cell line relative to the non-tumorigenic breast cell line MCF10A. In the first round of laboratory experiments GPT4 succeeded in discovering three drug combinations (out of 12 tested) with synergy scores above the positive controls. These combinations were itraconazole + atenolol, disulfiram + simvastatin and dipyridamole + mebendazole. GPT4 was then asked to generate new combinations after considering its initial results. It then discovered three more combinations with positive synergy scores (out of four tested), these were disulfiram + fulvestrant, mebendazole + quinacrine and disulfiram + quinacrine. A limitation of GPT4 as a generator of hypotheses was that its explanations for them were formulaic and unconvincing. We conclude that LLMs are an exciting novel source of scientific hypotheses.
Authors: Cailean Osborne, Jennifer Ding, Hannah Rose Kirk
Abstract: Open model developers have emerged as key actors in the political economy of artificial intelligence (AI), but we still have a limited understanding of collaborative practices in the open AI ecosystem. This paper responds to this gap with a three-part quantitative analysis of development activity on the Hugging Face (HF) Hub, a popular platform for building, sharing, and demonstrating models. First, various types of activity across 348,181 model, 65,761 dataset, and 156,642 space repositories exhibit right-skewed distributions. Activity is extremely imbalanced between repositories; for example, over 70% of models have 0 downloads, while 1% account for 99% of downloads. Furthermore, licenses matter: there are statistically significant differences in collaboration patterns in model repositories with permissive, restrictive, and no licenses. Second, we analyse a snapshot of the social network structure of collaboration in model repositories, finding that the community has a core-periphery structure, with a core of prolific developers and a majority of isolate developers (89%). Upon removing the isolate developers from the network, collaboration is characterised by high reciprocity regardless of developers' network positions. Third, we examine model adoption through the lens of model usage in spaces, finding that a minority of models, developed by a handful of companies, are widely used on the HF Hub. Overall, activity on the HF Hub is characterised by Pareto distributions, congruent with OSS development patterns on platforms like GitHub. We conclude with recommendations for researchers, companies, and policymakers to advance our understanding of open AI development.
Authors: Abrar Fahim, Alex Murphy, Alona Fyshe
Abstract: Multi-modal contrastive models such as CLIP achieve state-of-the-art performance in zero-shot classification by embedding input images and texts on a joint representational space. Recently, a modality gap has been reported in two-encoder contrastive models like CLIP, meaning that the image and text embeddings reside in disjoint areas of the latent space. Previous studies suggest that this gap exists due to 1) the cone effect, 2) mismatched pairs in the dataset, and 3) insufficient training. We show that, even when accounting for all these factors, and even when using the same modality, the contrastive loss actually creates a gap during training. As a result, We propose that the modality gap is inherent to the two-encoder contrastive loss and rename it the contrastive gap. We present evidence that attributes this contrastive gap to low uniformity in CLIP space, resulting in embeddings that occupy only a small portion of the latent space. To close the gap, we adapt the uniformity and alignment properties of unimodal contrastive loss to the multi-modal setting and show that simply adding these terms to the CLIP loss distributes the embeddings more uniformly in the representational space, closing the gap. In our experiments, we show that the modified representational space achieves better performance than default CLIP loss in downstream tasks such as zero-shot image classification and multi-modal arithmetic.
Authors: Matthew Watson, Divyashree Shivakumar Sreepathihalli, Francois Chollet, Martin Gorner, Kiranbir Sodhia, Ramesh Sampath, Tirth Patel, Haifeng Jin, Neel Kovelamudi, Gabriel Rasskin, Samaneh Saadat, Luke Wood, Chen Qian, Jonathan Bischof, Ian Stenbit, Abheesht Sharma, Anshuman Mishra
Abstract: We present the Keras domain packages KerasCV and KerasNLP, extensions of the Keras API for Computer Vision and Natural Language Processing workflows, capable of running on either JAX, TensorFlow, or PyTorch. These domain packages are designed to enable fast experimentation, with a focus on ease-of-use and performance. We adopt a modular, layered design: at the library's lowest level of abstraction, we provide building blocks for creating models and data preprocessing pipelines, and at the library's highest level of abstraction, we provide pretrained ``task" models for popular architectures such as Stable Diffusion, YOLOv8, GPT2, BERT, Mistral, CLIP, Gemma, T5, etc. Task models have built-in preprocessing, pretrained weights, and can be fine-tuned on raw inputs. To enable efficient training, we support XLA compilation for all models, and run all preprocessing via a compiled graph of TensorFlow operations using the tf.data API. The libraries are fully open-source (Apache 2.0 license) and available on GitHub.
Authors: Tianyang Xu, Shujin Wu, Shizhe Diao, Xiaoze Liu, Xingyao Wang, Yangyi Chen, Jing Gao
Abstract: Large language models (LLMs) often generate inaccurate or fabricated information and generally fail to indicate their confidence, which limits their broader applications. Previous work elicits confidence from LLMs by direct or self-consistency prompting, or constructing specific datasets for supervised finetuning. The prompting-based approaches have inferior performance, and the training-based approaches are limited to binary or inaccurate group-level confidence estimates. In this work, we present the advanced SaySelf, a training framework that teaches LLMs to express more accurate fine-grained confidence estimates. In addition, beyond the confidence scores, SaySelf initiates the process of directing LLMs to produce self-reflective rationales that clearly identify gaps in their parametric knowledge and explain their uncertainty. This is achieved by using an LLM to automatically summarize the uncertainties in specific knowledge via natural language. The summarization is based on the analysis of the inconsistency in multiple sampled reasoning chains, and the resulting data is utilized for supervised fine-tuning. Moreover, we utilize reinforcement learning with a meticulously crafted reward function to calibrate the confidence estimates, motivating LLMs to deliver accurate, high-confidence predictions and to penalize overconfidence in erroneous outputs. Experimental results in both in-distribution and out-of-distribution datasets demonstrate the effectiveness of SaySelf in reducing the confidence calibration error and maintaining the task performance. We show that the generated self-reflective rationales are reasonable and can further contribute to the calibration. The code is made public at https://github.com/xu1868/SaySelf.
Authors: Zhangchen Xu, Fengqing Jiang, Luyao Niu, Jinyuan Jia, Bo Li, Radha Poovendran
Abstract: In Federated Learning (FL), a set of clients collaboratively train a machine learning model (called global model) without sharing their local training data. The local training data of clients is typically non-i.i.d. and heterogeneous, resulting in varying contributions from individual clients to the final performance of the global model. In response, many contribution evaluation methods were proposed, where the server could evaluate the contribution made by each client and incentivize the high-contributing clients to sustain their long-term participation in FL. Existing studies mainly focus on developing new metrics or algorithms to better measure the contribution of each client. However, the security of contribution evaluation methods of FL operating in adversarial environments is largely unexplored. In this paper, we propose the first model poisoning attack on contribution evaluation methods in FL, termed ACE. Specifically, we show that any malicious client utilizing ACE could manipulate the parameters of its local model such that it is evaluated to have a high contribution by the server, even when its local training data is indeed of low quality. We perform both theoretical analysis and empirical evaluations of ACE. Theoretically, we show our design of ACE can effectively boost the malicious client's perceived contribution when the server employs the widely-used cosine distance metric to measure contribution. Empirically, our results show ACE effectively and efficiently deceive five state-of-the-art contribution evaluation methods. In addition, ACE preserves the accuracy of the final global models on testing inputs. We also explore six countermeasures to defend ACE. Our results show they are inadequate to thwart ACE, highlighting the urgent need for new defenses to safeguard the contribution evaluation methods in FL.
Authors: Nick Alonso, Tom\'as Figliolia, Anthony Ndirango, Beren Millidge
Abstract: There has recently been growing interest in conversational agents with long-term memory which has led to the rapid development of language models that use retrieval-augmented generation (RAG). Until recently, most work on RAG has focused on information retrieval from large databases of texts, like Wikipedia, rather than information from long-form conversations. In this paper, we argue that effective retrieval from long-form conversational data faces two unique problems compared to static database retrieval: 1) time/event-based queries, which requires the model to retrieve information about previous conversations based on time or the order of a conversational event (e.g., the third conversation on Tuesday), and 2) ambiguous queries that require surrounding conversational context to understand. To better develop RAG-based agents that can deal with these challenges, we generate a new dataset of ambiguous and time-based questions that build upon a recent dataset of long-form, simulated conversations, and demonstrate that standard RAG based approaches handle such questions poorly. We then develop a novel retrieval model which combines chained-of-table search methods, standard vector-database retrieval, and a prompting method to disambiguate queries, and demonstrate that this approach substantially improves over current methods at solving these tasks. We believe that this new dataset and more advanced RAG agent can act as a key benchmark and stepping stone towards effective memory augmented conversational agents that can be used in a wide variety of AI applications.
Authors: Yechen Xu, Xinhao Kong, Tingjun Chen, Danyang Zhuo
Abstract: The complexity of large language model (LLM) serving workloads has substantially increased due to the integration with external tool invocations, such as ChatGPT plugins. In this paper, we identify a new opportunity for efficient LLM serving for requests that trigger tools: tool partial execution alongside LLM decoding. To this end, we design Conveyor, an efficient LLM serving system optimized for handling requests involving external tools. We introduce a novel interface for tool developers to expose partial execution opportunities to the LLM serving system and a request scheduler that facilitates partial tool execution. Our results demonstrate that tool partial execution can improve request completion latency by up to 38.8%.
Authors: Lin Gui, Cristina G\^arbacea, Victor Veitch
Abstract: This paper concerns the problem of aligning samples from large language models to human preferences using best-of-$n$ sampling, where we draw $n$ samples, rank them, and return the best one. We consider two fundamental problems. First: what is the relationship between best-of-$n$ and approaches to alignment that train LLMs to output samples with a high expected reward (e.g., RLHF or DPO)? To answer this, we embed both the best-of-$n$ distribution and the sampling distributions learned by alignment procedures in a common class of tiltings of the base LLM distribution. We then show that, within this class, best-of-$n$ is essentially optimal in terms of the trade-off between win-rate against the base model vs KL distance from the base model. That is, best-of-$n$ is the best choice of alignment distribution if the goal is to maximize win rate. However, best-of-$n$ requires drawing $n$ samples for each inference, a substantial cost. To avoid this, the second problem we consider is how to fine-tune a LLM to mimic the best-of-$n$ sampling distribution. We derive BoNBoN Alignment to achieve this by exploiting the special structure of the best-of-$n$ distribution. Experiments show that BoNBoN alignment yields substantial improvements in producing a model that is preferred to the base policy while minimally affecting off-target aspects.
Authors: Zicheng Liu, Jiahui Li, Siyuan Li, Zelin Zang, Cheng Tan, Yufei Huang, Yajing Bai, Stan Z. Li
Abstract: The Genomic Foundation Model (GFM) paradigm is expected to facilitate the extraction of generalizable representations from massive genomic data, thereby enabling their application across a spectrum of downstream applications. Despite advancements, a lack of evaluation framework makes it difficult to ensure equitable assessment due to experimental settings, model intricacy, benchmark datasets, and reproducibility challenges. In the absence of standardization, comparative analyses risk becoming biased and unreliable. To surmount this impasse, we introduce GenBench, a comprehensive benchmarking suite specifically tailored for evaluating the efficacy of Genomic Foundation Models. GenBench offers a modular and expandable framework that encapsulates a variety of state-of-the-art methodologies. Through systematic evaluations of datasets spanning diverse biological domains with a particular emphasis on both short-range and long-range genomic tasks, firstly including the three most important DNA tasks covering Coding Region, Non-Coding Region, Genome Structure, etc. Moreover, We provide a nuanced analysis of the interplay between model architecture and dataset characteristics on task-specific performance. Our findings reveal an interesting observation: independent of the number of parameters, the discernible difference in preference between the attention-based and convolution-based models on short- and long-range tasks may provide insights into the future design of GFM.
Authors: Qian Lou, Xin Liang, Jiaqi Xue, Yancheng Zhang, Rui Xie, Mengxin Zheng
Abstract: It is imperative to ensure the stability of every prediction made by a language model; that is, a language's prediction should remain consistent despite minor input variations, like word substitutions. In this paper, we investigate the problem of certifying a language model's robustness against Universal Text Perturbations (UTPs), which have been widely used in universal adversarial attacks and backdoor attacks. Existing certified robustness based on random smoothing has shown considerable promise in certifying the input-specific text perturbations (ISTPs), operating under the assumption that any random alteration of a sample's clean or adversarial words would negate the impact of sample-wise perturbations. However, with UTPs, masking only the adversarial words can eliminate the attack. A naive method is to simply increase the masking ratio and the likelihood of masking attack tokens, but it leads to a significant reduction in both certified accuracy and the certified radius due to input corruption by extensive masking. To solve this challenge, we introduce a novel approach, the superior prompt search method, designed to identify a superior prompt that maintains higher certified accuracy under extensive masking. Additionally, we theoretically motivate why ensembles are a particularly suitable choice as base prompts for random smoothing. The method is denoted by superior prompt ensembling technique. We also empirically confirm this technique, obtaining state-of-the-art results in multiple settings. These methodologies, for the first time, enable high certified accuracy against both UTPs and ISTPs. The source code of CR-UTP is available at \url {https://github.com/UCFML-Research/CR-UTP}.
Authors: Ziyun Liang, Xiaoqing Guo, J. Alison Noble, Konstantinos Kamnitsas
Abstract: Unsupervised anomaly segmentation approaches to pathology segmentation train a model on images of healthy subjects, that they define as the 'normal' data distribution. At inference, they aim to segment any pathologies in new images as 'anomalies', as they exhibit patterns that deviate from those in 'normal' training data. Prevailing methods follow the 'corrupt-and-reconstruct' paradigm. They intentionally corrupt an input image, reconstruct it to follow the learned 'normal' distribution, and subsequently segment anomalies based on reconstruction error. Corrupting an input image, however, inevitably leads to suboptimal reconstruction even of normal regions, causing false positives. To alleviate this, we propose a novel iterative spatial mask-refining strategy IterMask2. We iteratively mask areas of the image, reconstruct them, and update the mask based on reconstruction error. This iterative process progressively adds information about areas that are confidently normal as per the model. The increasing content guides reconstruction of nearby masked areas, improving reconstruction of normal tissue under these areas, reducing false positives. We also use high-frequency image content as an auxiliary input to provide additional structural information for masked areas. This further improves reconstruction error of normal in comparison to anomalous areas, facilitating segmentation of the latter. We conduct experiments on several brain lesion datasets and demonstrate effectiveness of our method. Code is available at: https://github.com/ZiyunLiang/IterMask2