Authors: Tim Z. Xiao, Johannes Zenn, Robert Bamler
The Street View House Numbers (SVHN) dataset is a popular benchmark dataset in deep learning. Originally designed for digit classification tasks, the SVHN dataset has been widely used as a benchmark for various other tasks including generative modeling. However, with this work, we aim to warn the community about an issue of the SVHN dataset as a benchmark for generative modeling tasks: we discover that the official split into training set and test set of the SVHN dataset are not drawn from the same distribution. We empirically show that this distribution mismatch has little impact on the classification task (which may explain why this issue has not been detected before), but it severely affects the evaluation of probabilistic generative models, such as Variational Autoencoders and diffusion models. As a workaround, we propose to mix and re-split the official training and test set when SVHN is used for tasks other than classification. We publish a new split and the indices we used to create it at https://jzenn.github.io/svhn-remix/ .
Authors: Ryan M. Dreifuerst, Robert W. Heath Jr
Beam management is a strategy to unify beamforming and channel state information (CSI) acquisition with large antenna arrays in 5G. Codebooks serve multiple uses in beam management including beamforming reference signals, CSI reporting, and analog beam training. In this paper, we propose and evaluate a machine learning-refined codebook design process for extremely large multiple-input multiple-output (X-MIMO) systems. We propose a neural network and beam selection strategy to design the initial access and refinement codebooks using end-to-end learning from beamspace representations. The algorithm, called Extreme-Beam Management (X-BM), can significantly improve the performance of extremely large arrays as envisioned for 6G and capture realistic wireless and physical layer aspects. Our results show an 8dB improvement in initial access and overall effective spectral efficiency improvements compared to traditional codebook methods.
Authors: Du Phan, Matthew D. Hoffman, David Dohan, Sholto Douglas, Tuan Anh Le, Aaron Parisi, Pavel Sountsov, Charles Sutton, Sharad Vikram, Rif A. Saurous
Large language models (LLMs) solve problems more accurately and interpretably when instructed to work out the answer step by step using a ``chain-of-thought'' (CoT) prompt. One can also improve LLMs' performance on a specific task by supervised fine-tuning, i.e., by using gradient ascent on some tunable parameters to maximize the average log-likelihood of correct answers from a labeled training set. Naively combining CoT with supervised tuning requires supervision not just of the correct answers, but also of detailed rationales that lead to those answers; these rationales are expensive to produce by hand. Instead, we propose a fine-tuning strategy that tries to maximize the \emph{marginal} log-likelihood of generating a correct answer using CoT prompting, approximately averaging over all possible rationales. The core challenge is sampling from the posterior over rationales conditioned on the correct answer; we address it using a simple Markov-chain Monte Carlo (MCMC) expectation-maximization (EM) algorithm inspired by the self-taught reasoner (STaR), memoized wake-sleep, Markovian score climbing, and persistent contrastive divergence. This algorithm also admits a novel control-variate technique that drives the variance of our gradient estimates to zero as the model improves. Applying our technique to GSM8K and the tasks in BIG-Bench Hard, we find that this MCMC-EM fine-tuning technique typically improves the model's accuracy on held-out examples more than STaR or prompt-tuning with or without CoT.
Authors: Zeteng Lin
With the rapid development of artificial intelligence and breakthroughs in machine learning and natural language processing, intelligent question-answering robots have become widely used in government affairs. This paper conducts a horizontal comparison between Guangdong Province's government chatbots, ChatGPT, and Wenxin Ernie, two large language models, to analyze the strengths and weaknesses of existing government chatbots and AIGC technology. The study finds significant differences between government chatbots and large language models. China's government chatbots are still in an exploratory stage and have a gap to close to achieve "intelligence." To explore the future direction of government chatbots more deeply, this research proposes targeted optimization paths to help generative AI be effectively applied in government chatbot conversations.
Authors: Keisuke Suzuki
In this paper, we prove that an Adam-type algorithm with smooth clipping approaches the global minimizer of the regularized non-convex loss function. Adding smooth clipping and taking the state space as the set of all trajectories, we can apply the ergodic theory of Markov semigroups for this algorithm and investigate its asymptotic behavior. The ergodic theory we establish in this paper reduces the problem of evaluating the convergence, generalization error and discretization error of this algorithm to the problem of evaluating the difference between two functional stochastic differential equations (SDEs) with different drift coefficients. As a result of our analysis, we have shown that this algorithm minimizes the the regularized non-convex loss function with errors of the form $n^{-1/2}$, $\eta^{1/4}$, $\beta^{-1} \log (\beta + 1)$ and $e^{- c t}$. Here, $c$ is a constant and $n$, $\eta$, $\beta$ and $t$ denote the size of the training dataset, learning rate, inverse temperature and time, respectively.
Authors: Jiwei Zhao, Jiacheng Chen, Zeyu Sun, Yuhang Shi, Haibo Zhou, Xuemin (Sherman) Shen
As the demand for high-quality services proliferates, an innovative network architecture, the fully-decoupled RAN (FD-RAN), has emerged for more flexible spectrum resource utilization and lower network costs. However, with the decoupling of uplink base stations and downlink base stations in FD-RAN, the traditional transmission mechanism, which relies on real-time channel feedback, is not suitable as the receiver is not able to feedback accurate and timely channel state information to the transmitter. This paper proposes a novel transmission scheme without relying on physical layer channel feedback. Specifically, we design a radio map based complex-valued precoding network~(RMCPNet) model, which outputs the base station precoding based on user location. RMCPNet comprises multiple subnets, with each subnet responsible for extracting unique modal features from diverse input modalities. Furthermore, the multi-modal embeddings derived from these distinct subnets are integrated within the information fusion layer, culminating in a unified representation. We also develop a specific RMCPNet training algorithm that employs the negative spectral efficiency as the loss function. We evaluate the performance of the proposed scheme on the public DeepMIMO dataset and show that RMCPNet can achieve 16\% and 76\% performance improvements over the conventional real-valued neural network and statistical codebook approach, respectively.
Authors: Duc-Anh Nguyen, Cuong Pham, Nhien-An Le-Khac
Various types of sensors can be used for Human Activity Recognition (HAR), and each of them has different strengths and weaknesses. Sometimes a single sensor cannot fully observe the user's motions from its perspective, which causes wrong predictions. While sensor fusion provides more information for HAR, it comes with many inherent drawbacks like user privacy and acceptance, costly set-up, operation, and maintenance. To deal with this problem, we propose Virtual Fusion - a new method that takes advantage of unlabeled data from multiple time-synchronized sensors during training, but only needs one sensor for inference. Contrastive learning is adopted to exploit the correlation among sensors. Virtual Fusion gives significantly better accuracy than training with the same single sensor, and in some cases, it even surpasses actual fusion using multiple sensors at test time. We also extend this method to a more general version called Actual Fusion within Virtual Fusion (AFVF), which uses a subset of training sensors during inference. Our method achieves state-of-the-art accuracy and F1-score on UCI-HAR and PAMAP2 benchmark datasets. Implementation is available upon request.
Authors: Joseph Paul Cohen, Louis Blankemeier, Akshay Chaudhari
Models driven by spurious correlations often yield poor generalization performance. We propose the counterfactual alignment method to detect and explore spurious correlations of black box classifiers. Counterfactual images generated with respect to one classifier can be input into other classifiers to see if they also induce changes in the outputs of these classifiers. The relationship between these responses can be quantified and used to identify specific instances where a spurious correlation exists as well as compute aggregate statistics over a dataset. Our work demonstrates the ability to detect spurious correlations in face attribute classifiers. This is validated by observing intuitive trends in a face attribute classifier as well as fabricating spurious correlations and detecting their presence, both visually and quantitatively. Further, utilizing the CF alignment method, we demonstrate that we can rectify spurious correlations identified in classifiers.
Authors: Mark Peelen, Leila Bagheriye, Johan Kwisthout
The integration of multi-omics data has emerged as a promising approach for gaining comprehensive insights into complex diseases such as cancer. This paper proposes a novel approach to identify cancer subtypes through the integration of multi-omics data for clustering. The proposed method, named LIDAF utilises affinity matrices based on linear relationships between and within different omics datasets (Linear Inter and Intra Dataset Affinity Fusion (LIDAF)). Canonical Correlation Analysis is in this paper employed to create distance matrices based on Euclidean distances between canonical variates. The distance matrices are converted to affinity matrices and those are fused in a three-step process. The proposed LIDAF addresses the limitations of the existing method resulting in improvement of clustering performance as measured by the Adjusted Rand Index and the Normalized Mutual Information score. Moreover, our proposed LIDAF approach demonstrates a notable enhancement in 50% of the log10 rank p-values obtained from Cox survival analysis, surpassing the performance of the best reported method, highlighting its potential of identifying distinct cancer subtypes.
Authors: Jeremy Irvin, Lucas Tao, Joanne Zhou, Yuntao Ma, Langston Nashold, Benjamin Liu, Andrew Y. Ng
Large, self-supervised vision models have led to substantial advancements for automatically interpreting natural images. Recent works have begun tailoring these methods to remote sensing data which has rich structure with multi-sensor, multi-spectral, and temporal information providing massive amounts of self-labeled data that can be used for self-supervised pre-training. In this work, we develop a new encoder architecture called USat that can input multi-spectral data from multiple sensors for self-supervised pre-training. USat is a vision transformer with modified patch projection layers and positional encodings to model spectral bands with varying spatial scales from multiple sensors. We integrate USat into a Masked Autoencoder (MAE) self-supervised pre-training procedure and find that a pre-trained USat outperforms state-of-the-art self-supervised MAE models trained on remote sensing data on multiple remote sensing benchmark datasets (up to 8%) and leads to improvements in low data regimes (up to 7%). Code and pre-trained weights are available at https://github.com/stanfordmlgroup/USat .
Authors: Weikang Qiu, Huangrui Chu, Selena Wang, Haolan Zuo, Xiaoxiao Li, Yize Zhao, Rex Ying
Discovering reliable and informative interactions among brain regions from functional magnetic resonance imaging (fMRI) signals is essential in neuroscientific predictions of cognition. Most of the current methods fail to accurately characterize those interactions because they only focus on pairwise connections and overlook the high-order relationships of brain regions. We delve into this problem and argue that these high-order relationships should be maximally informative and minimally redundant (MIMR). However, identifying such high-order relationships is challenging and highly under-explored. Methods that can be tailored to our context are also non-existent. In response to this gap, we propose a novel method named HyBRiD that aims to extract MIMR high-order relationships from fMRI data. HyBRiD employs a Constructor to identify hyperedge structures, and a Weighter to compute a weight for each hyperedge. HyBRiD achieves the MIMR objective through an innovative information bottleneck framework named multi-head drop-bottleneck with theoretical guarantees. Our comprehensive experiments demonstrate the effectiveness of our model. Our model outperforms the state-of-the-art predictive model by an average of 12.1%, regarding the quality of hyperedges measured by CPM, a standard protocol for studying brain connections.
Authors: Charles-Étienne Joseph, Benjamin Thérien, Abhinav Moudgil, Boris Knyazev, Eugene Belilovsky
Communication-efficient variants of SGD, specifically local SGD, have received a great deal of interest in recent years. These approaches compute multiple gradient steps locally, that is on each worker, before averaging model parameters, helping relieve the critical communication bottleneck in distributed deep learning training. Although many variants of these approaches have been proposed, they can sometimes lag behind state-of-the-art adaptive optimizers for deep learning. In this work, we investigate if the recent progress in the emerging area of learned optimizers can potentially close this gap while remaining communication-efficient. Specifically, we meta-learn how to perform global updates given an update from local SGD iterations. Our results demonstrate that learned optimizers can substantially outperform local SGD and its sophisticated variants while maintaining their communication efficiency. Learned optimizers can even generalize to unseen and much larger datasets and architectures, including ImageNet and ViTs, and to unseen modalities such as language modeling. We therefore demonstrate the potential of learned optimizers for improving communication-efficient distributed learning.
Authors: Neha Kalibhat, Warren Morningstar, Alex Bijamov, Luyang Liu, Karan Singhal, Philip Mansfield
Self-Supervised Learning (SSL) enables training performant models using limited labeled data. One of the pillars underlying vision SSL is the use of data augmentations/perturbations of the input which do not significantly alter its semantic content. For audio and other temporal signals, augmentations are commonly used alongside format transforms such as Fourier transforms or wavelet transforms. Unlike augmentations, format transforms do not change the information contained in the data; rather, they express the same information in different coordinates. In this paper, we study the effects of format transforms and augmentations both separately and together on vision SSL. We define augmentations in frequency space called Fourier Domain Augmentations (FDA) and show that training SSL models on a combination of these and image augmentations can improve the downstream classification accuracy by up to 1.3% on ImageNet-1K. We also show improvements against SSL baselines in few-shot and transfer learning setups using FDA. Surprisingly, we also observe that format transforms can improve the quality of learned representations even without augmentations; however, the combination of the two techniques yields better quality.
Authors: Seyedarmin Azizi, Mahdi Nazemi, Mehdi Kamal, Massoud Pedram
This paper presents a mixed-computation neural network processing approach for edge applications that incorporates low-precision (low-width) Posit and low-precision fixed point (FixP) number systems. This mixed-computation approach employs 4-bit Posit (Posit4), which has higher precision around zero, for representing weights with high sensitivity, while it uses 4-bit FixP (FixP4) for representing other weights. A heuristic for analyzing the importance and the quantization error of the weights is presented to assign the proper number system to different weights. Additionally, a gradient approximation for Posit representation is introduced to improve the quality of weight updates in the backpropagation process. Due to the high energy consumption of the fully Posit-based computations, neural network operations are carried out in FixP or Posit/FixP. An efficient hardware implementation of a MAC operation with a first Posit operand and FixP for a second operand and accumulator is presented. The efficacy of the proposed low-precision mixed-computation approach is extensively assessed on vision and language models. The results show that, on average, the accuracy of the mixed-computation is about 1.5% higher than that of FixP with a cost of 0.19% energy overhead.
Authors: Shang-Ching Liu, ShengKun Wang, Wenqi Lin, Chung-Wei Hsiung, Yi-Chen Hsieh, Yu-Ping Cheng, Sian-Hong Luo, Tsungyao Chang, Jianwei Zhang
In this study, we introduce JarviX, a sophisticated data analytics framework. JarviX is designed to employ Large Language Models (LLMs) to facilitate an automated guide and execute high-precision data analyzes on tabular datasets. This framework emphasizes the significance of varying column types, capitalizing on state-of-the-art LLMs to generate concise data insight summaries, propose relevant analysis inquiries, visualize data effectively, and provide comprehensive explanations for results drawn from an extensive data analysis pipeline. Moreover, JarviX incorporates an automated machine learning (AutoML) pipeline for predictive modeling. This integration forms a comprehensive and automated optimization cycle, which proves particularly advantageous for optimizing machine configuration. The efficacy and adaptability of JarviX are substantiated through a series of practical use case studies.
Authors: Amit Baras, Alon Zolfi, Yuval Elovici, Asaf Shabtai
In recent years, there has been a significant trend in deep neural networks (DNNs), particularly transformer-based models, of developing ever-larger and more capable models. While they demonstrate state-of-the-art performance, their growing scale requires increased computational resources (e.g., GPUs with greater memory capacity). To address this problem, quantization techniques (i.e., low-bit-precision representation and matrix multiplication) have been proposed. Most quantization techniques employ a static strategy in which the model parameters are quantized, either during training or inference, without considering the test-time sample. In contrast, dynamic quantization techniques, which have become increasingly popular, adapt during inference based on the input provided, while maintaining full-precision performance. However, their dynamic behavior and average-case performance assumption makes them vulnerable to a novel threat vector -- adversarial attacks that target the model's efficiency and availability. In this paper, we present QuantAttack, a novel attack that targets the availability of quantized models, slowing down the inference, and increasing memory usage and energy consumption. We show that carefully crafted adversarial examples, which are designed to exhaust the resources of the operating system, can trigger worst-case performance. In our experiments, we demonstrate the effectiveness of our attack on vision transformers on a wide range of tasks, both uni-modal and multi-modal. We also examine the effect of different attack variants (e.g., a universal perturbation) and the transferability between different models.
Authors: Alessandro Farace di Villaforesta, Lucie Charlotte Magister, Pietro Barbiero, Pietro Liò
To address the challenge of the ``black-box" nature of deep learning in medical settings, we combine GCExplainer - an automated concept discovery solution - along with Logic Explained Networks to provide global explanations for Graph Neural Networks. We demonstrate this using a generally applicable graph construction and classification pipeline, involving panoptic segmentation with HoVer-Net and cancer prediction with Graph Convolution Networks. By training on H&E slides of breast cancer, we show promising results in offering explainable and trustworthy AI tools for clinicians.
Authors: Cong-Duy Nguyen, Thong Nguyen, Duc Anh Vu, Luu Anh Tuan
The effectiveness of a model is heavily reliant on the quality of the fusion representation of multiple modalities in multimodal sentiment analysis. Moreover, each modality is extracted from raw input and integrated with the rest to construct a multimodal representation. Although previous methods have proposed multimodal representations and achieved promising results, most of them focus on forming positive and negative pairs, neglecting the variation in sentiment scores within the same class. Additionally, they fail to capture the significance of unimodal representations in the fusion vector. To address these limitations, we introduce a framework called Supervised Angular-based Contrastive Learning for Multimodal Sentiment Analysis. This framework aims to enhance discrimination and generalizability of the multimodal representation and overcome biases in the fusion vector's modality. Our experimental results, along with visualizations on two widely used datasets, demonstrate the effectiveness of our approach.
Authors: Mahboobeh Parsapoor
Changes in speech and language are among the first signs of Parkinson's disease (PD). Thus, clinicians have tried to identify individuals with PD from their voices for years. Doctors can leverage AI-based speech assessments to spot PD thanks to advancements in artificial intelligence (AI). Such AI systems can be developed using machine learning classifiers that have been trained using individuals' voices. Although several studies have shown reasonable results in developing such AI systems, these systems would need more data samples to achieve promising performance. This paper explores using deep learning-based data generation techniques on the accuracy of machine learning classifiers that are the core of such systems.
Authors: Yunhui Jang, Seul Lee, Sungsoo Ahn
Recently, there has been a surge of interest in employing neural networks for graph generation, a fundamental statistical learning problem with critical applications like molecule design and community analysis. However, most approaches encounter significant limitations when generating large-scale graphs. This is due to their requirement to output the full adjacency matrices whose size grows quadratically with the number of nodes. In response to this challenge, we introduce a new, simple, and scalable graph representation named gap encoded edge list (GEEL) that has a small representation size that aligns with the number of edges. In addition, GEEL significantly reduces the vocabulary size by incorporating the gap encoding and bandwidth restriction schemes. GEEL can be autoregressively generated with the incorporation of node positional encoding, and we further extend GEEL to deal with attributed graphs by designing a new grammar. Our findings reveal that the adoption of this compact representation not only enhances scalability but also bolsters performance by simplifying the graph generation process. We conduct a comprehensive evaluation across ten non-attributed and two molecular graph generation tasks, demonstrating the effectiveness of GEEL.
Authors: Guanlin Li, Han Qiu, Shangwei Guo, Jiwei Li, Tianwei Zhang
Adversarial training (AT) is an important and attractive topic in deep learning security, exhibiting mysteries and odd properties. Recent studies of neural network training dynamics based on Neural Tangent Kernel (NTK) make it possible to reacquaint AT and deeply analyze its properties. In this paper, we perform an in-depth investigation of AT process and properties with NTK, such as NTK evolution. We uncover three new findings that are missed in previous works. First, we disclose the impact of data normalization on AT and the importance of unbiased estimators in batch normalization layers. Second, we experimentally explore the kernel dynamics and propose more time-saving AT methods. Third, we study the spectrum feature inside the kernel to address the catastrophic overfitting problem. To the best of our knowledge, it is the first work leveraging the observations of kernel dynamics to improve existing AT methods.
Authors: Taha Yassine (IETR, INSA Rennes), Baptiste Chatelier (IETR, MERCE-France, INSA Rennes), Vincent Corlay (MERCE-France), Matthieu Crussière (IETR, INSA Rennes), Stephane Paquelet, Olav Tirkkonen, Luc Le Magoarou (INSA Rennes, IETR)
Channel charting builds a map of the radio environment in an unsupervised way. The obtained chart locations can be seen as low-dimensional compressed versions of channel state information that can be used for a wide variety of applications, including beam prediction. In non-standalone or cell-free systems, chart locations computed at a given base station can be transmitted to several other base stations (possibly operating at different frequency bands) for them to predict which beams to use. This potentially yields a dramatic reduction of the overhead due to channel estimation or beam management, since only the base station performing charting requires channel state information, the others directly predicting the beam from the chart location. In this paper, advanced model-based neural network architectures are proposed for both channel charting and beam prediction. The proposed methods are assessed on realistic synthetic channels, yielding promising results.
Authors: Nan Chen, Zhihong Li, Jun Tao
Flow fields are often partitioned into data blocks for massively parallel computation and analysis based on blockwise relationships. However, most of the previous techniques only consider the first-order dependencies among blocks, which is insufficient in describing complex flow patterns. In this work, we present FlowHON, an approach to construct higher-order networks (HONs) from flow fields. FlowHON captures the inherent higher-order dependencies in flow fields as nodes and estimates the transitions among them as edges. We formulate the HON construction as an optimization problem with three linear transformations. The first two layers correspond to the node generation and the third one corresponds to edge estimation. Our formulation allows the node generation and edge estimation to be solved in a unified framework. With FlowHON, the rich set of traditional graph algorithms can be applied without any modification to analyze flow fields, while leveraging the higher-order information to understand the inherent structure and manage flow data for efficiency. We demonstrate the effectiveness of FlowHON using a series of downstream tasks, including estimating the density of particles during tracing, partitioning flow fields for data management, and understanding flow fields using the node-link diagram representation of networks.
Authors: Gabriel della Maggiora, Luis Alberto Croquevielle, Nikita Desphande, Harry Horsley, Thomas Heinis, Artur Yakimovich
Inverse problems aim to determine parameters from observations, a crucial task in engineering and science. Lately, generative models, especially diffusion models, have gained popularity in this area for their ability to produce realistic solutions and their good mathematical properties. Despite their success, an important drawback of diffusion models is their sensitivity to the choice of variance schedule, which controls the dynamics of the diffusion process. Fine-tuning this schedule for specific applications is crucial but time-costly and does not guarantee an optimal result. We propose a novel approach for learning the schedule as part of the training process. Our method supports probabilistic conditioning on data, provides high-quality solutions, and is flexible, proving able to adapt to different applications with minimum overhead. This approach is tested in two unrelated inverse problems: super-resolution microscopy and quantitative phase imaging, yielding comparable or superior results to previous methods and fine-tuned diffusion models. We conclude that fine-tuning the schedule by experimentation should be avoided because it can be learned during training in a stable way that yields better results.
Authors: Razvan Caramalau, Binod Bhattarai, Danail Stoyanov
In this paper, we introduce Active Learning framework in Federated Learning for Target Domain Generalisation, harnessing the strength from both learning paradigms. Our framework, FEDALV, composed of Active Learning (AL) and Federated Domain Generalisation (FDG), enables generalisation of an image classification model trained from limited source domain client's data without sharing images to an unseen target domain. To this end, our FDG, FEDA, consists of two optimisation updates during training, one at the client and another at the server level. For the client, the introduced losses aim to reduce feature complexity and condition alignment, while in the server, the regularisation limits free energy biases between source and target obtained by the global model. The remaining component of FEDAL is AL with variable budgets, which queries the server to retrieve and sample the most informative local data for the targeted client. We performed multiple experiments on FDG w/ and w/o AL and compared with both conventional FDG baselines and Federated Active Learning baselines. Our extensive quantitative experiments demonstrate the superiority of our method in accuracy and efficiency compared to the multiple contemporary methods. FEDALV manages to obtain the performance of the full training target accuracy while sampling as little as 5% of the source client's data.
Authors: Sophia Krix, Ella Wilczynski, Neus Falgàs, Raquel Sánchez-Valle, Eti Yoles, Uri Nevo, Kuti Baruch, Holger Fröhlich
Alzheimer's disease has an increasing prevalence in the population world-wide, yet current diagnostic methods based on recommended biomarkers are only available in specialized clinics. Due to these circumstances, Alzheimer's disease is usually diagnosed late, which contrasts with the currently available treatment options that are only effective for patients at an early stage. Blood-based biomarkers could fill in the gap of easily accessible and low-cost methods for early diagnosis of the disease. In particular, immune-based blood-biomarkers might be a promising option, given the recently discovered cross-talk of immune cells of the central nervous system with those in the peripheral immune system. With the help of machine learning algorithms and mechanistic modeling approaches, such as agent-based modeling, an in-depth analysis of the simulation of cell dynamics is possible as well as of high-dimensional omics resources indicative of pathway signaling changes. Here, we give a background on advances in research on brain-immune system cross-talk in Alzheimer's disease and review recent machine learning and mechanistic modeling approaches which leverage modern omics technologies for blood-based immune system-related biomarker discovery.
Authors: Amine Rebei
The ability to generate SQL queries from natural language has significant implications for making data accessible to non-specialists. This paper presents a novel approach to fine-tuning open-source large language models (LLMs) for the task of transforming natural language into SQL queries within the retail domain. We introduce models specialized in generating SQL queries, trained on synthetic datasets tailored to the Snowflake SQL and GoogleSQL dialects. Our methodology involves generating a context-specific dataset using GPT-4, then fine-tuning three open-source LLMs(Starcoder Plus, Code-Llama, and Mistral) employing the LoRa technique to optimize for resource constraints. The fine-tuned models demonstrate superior performance in zero-shot settings compared to the baseline GPT-4, with Code-Llama achieving the highest accuracy rates, at 81.58% for Snowflake SQL and 82.66% for GoogleSQL. These results underscore the effectiveness of fine-tuning LLMs on domain-specific tasks and suggest a promising direction for enhancing the accessibility of relational databases through natural language interfaces.
Authors: Zhuoran Yu, Chenchen Zhu, Sean Culatana, Raghuraman Krishnamoorthi, Fanyi Xiao, Yong Jae Lee
Recent advances in generative deep learning have enabled the creation of high-quality synthetic images in text-to-image generation. Prior work shows that fine-tuning a pretrained diffusion model on ImageNet and generating synthetic training images from the finetuned model can enhance an ImageNet classifier's performance. However, performance degrades as synthetic images outnumber real ones. In this paper, we explore whether generative fine-tuning is essential for this improvement and whether it is possible to further scale up training using more synthetic data. We present a new framework leveraging off-the-shelf generative models to generate synthetic training images, addressing multiple challenges: class name ambiguity, lack of diversity in naive prompts, and domain shifts. Specifically, we leverage large language models (LLMs) and CLIP to resolve class name ambiguity. To diversify images, we propose contextualized diversification (CD) and stylized diversification (SD) methods, also prompted by LLMs. Finally, to mitigate domain shifts, we leverage domain adaptation techniques with auxiliary batch normalization for synthetic images. Our framework consistently enhances recognition model performance with more synthetic data, up to 6x of original ImageNet size showcasing the potential of synthetic data for improved recognition models and strong out-of-domain generalization.
Authors: Ishaan Gupta, Samyutha Ayalasomayajula, Yashas Shashidhara, Anish Kataria, Shreyas Shashidhara, Krishita Kataria, Aditya Undurti
The prediction of crop yields internationally is a crucial objective in agricultural research. Thus, this study implements 6 regression models (Linear, Tree, Gradient Descent, Gradient Boosting, K- Nearest Neighbors, and Random Forest) to predict crop yields in 196 countries. Given 4 key training parameters, pesticides (tonnes), rainfall (mm), temperature (Celsius), and yield (hg/ha), it was found that our Random Forest Regression model achieved a determination coefficient (r^2) of 0.94, with a margin of error (ME) of .03. The models were trained and tested using the Food and Agricultural Organization of the United Nations data, along with the World Bank Climate Change Data Catalog. Furthermore, each parameter was analyzed to understand how varying factors could impact overall yield. We used unconventional models, contrary to generally used Deep Learning (DL) and Machine Learning (ML) models, combined with recently collected data to implement a unique approach in our research. Existing scholarship would benefit from understanding the most optimal model for agricultural research, specifically using the United Nations data.
Authors: Felix Tristram, Stefano Gasperini, Federico Tombari, Nassir Navab, Benjamin Busam
Neural Radiance Fields (NeRFs) have shown remarkable novel view synthesis capabilities even in large-scale, unbounded scenes, albeit requiring hundreds of views or introducing artifacts in sparser settings. Their optimization suffers from shape-radiance ambiguities wherever only a small visual overlap is available. This leads to erroneous scene geometry and artifacts. In this paper, we propose Re-Nerfing, a simple and general multi-stage approach that leverages NeRF's own view synthesis to address these limitations. With Re-Nerfing, we increase the scene's coverage and enhance the geometric consistency of novel views as follows: First, we train a NeRF with the available views. Then, we use the optimized NeRF to synthesize pseudo-views next to the original ones to simulate a stereo or trifocal setup. Finally, we train a second NeRF with both original and pseudo views while enforcing structural, epipolar constraints via the newly synthesized images. Extensive experiments on the mip-NeRF 360 dataset show the effectiveness of Re-Nerfing across denser and sparser input scenarios, bringing improvements to the state-of-the-art Zip-NeRF, even when trained with all views.
Authors: Joshua Batson, Yonatan Kahn
We demonstrate the emergence of scaling laws in the benchmark top versus QCD jet classification problem in collider physics. Six distinct physically-motivated classifiers exhibit power-law scaling of the binary cross-entropy test loss as a function of training set size, with distinct power law indices. This result highlights the importance of comparing classifiers as a function of dataset size rather than for a fixed training set, as the optimal classifier may change considerably as the dataset is scaled up. We speculate on the interpretation of our results in terms of previous models of scaling laws observed in natural language and image datasets.
Authors: Bokun Wang, Tianbao Yang
This paper revisits a class of convex Finite-Sum Coupled Compositional Stochastic Optimization (cFCCO) problems with many applications, including group distributionally robust optimization (GDRO), reinforcement learning, and learning to rank. To better solve these problems, we introduce a unified family of efficient single-loop primal-dual block-coordinate proximal algorithms, dubbed ALEXR. This algorithm leverages block-coordinate stochastic mirror ascent updates for the dual variable and stochastic proximal gradient descent updates for the primal variable. We establish the convergence rates of ALEXR in both convex and strongly convex cases under smoothness and non-smoothness conditions of involved functions, which not only improve the best rates in previous works on smooth cFCCO problems but also expand the realm of cFCCO for solving more challenging non-smooth problems such as the dual form of GDRO. Finally, we present lower complexity bounds to demonstrate that the convergence rates of ALEXR are optimal among first-order block-coordinate stochastic algorithms for the considered class of cFCCO problems.
Authors: Akshay Goel, Almog Gueta, Omry Gilon, Chang Liu, Sofia Erell, Lan Huong Nguyen, Xiaohong Hao, Bolous Jaber, Shashir Reddy, Rupesh Kartha, Jean Steiner, Itay Laish, Amir Feder
The unstructured nature of clinical notes within electronic health records often conceals vital patient-related information, making it challenging to access or interpret. To uncover this hidden information, specialized Natural Language Processing (NLP) models are required. However, training these models necessitates large amounts of labeled data, a process that is both time-consuming and costly when relying solely on human experts for annotation. In this paper, we propose an approach that combines Large Language Models (LLMs) with human expertise to create an efficient method for generating ground truth labels for medical text annotation. By utilizing LLMs in conjunction with human annotators, we significantly reduce the human annotation burden, enabling the rapid creation of labeled datasets. We rigorously evaluate our method on a medical information extraction task, demonstrating that our approach not only substantially cuts down on human intervention but also maintains high accuracy. The results highlight the potential of using LLMs to improve the utilization of unstructured clinical data, allowing for the swift deployment of tailored NLP solutions in healthcare.
Authors: Jiaxin Gao, Qinglong Cao, Yuntian Chen
Automatic Modulation Classification (AMC) plays a vital role in time series analysis, such as signal classification and identification within wireless communications. Deep learning-based AMC models have demonstrated significant potential in this domain. However, current AMC models inadequately consider the disparities in handling signals under conditions of low and high Signal-to-Noise Ratio (SNR), resulting in an unevenness in their performance. In this study, we propose MoE-AMC, a novel Mixture-of-Experts (MoE) based model specifically crafted to address AMC in a well-balanced manner across varying SNR conditions. Utilizing the MoE framework, MoE-AMC seamlessly combines the strengths of LSRM (a Transformer-based model) for handling low SNR signals and HSRM (a ResNet-based model) for high SNR signals. This integration empowers MoE-AMC to achieve leading performance in modulation classification, showcasing its efficacy in capturing distinctive signal features under diverse SNR scenarios. We conducted experiments using the RML2018.01a dataset, where MoE-AMC achieved an average classification accuracy of 71.76% across different SNR levels, surpassing the performance of previous SOTA models by nearly 10%. This study represents a pioneering application of MoE techniques in the realm of AMC, offering a promising avenue for elevating signal classification accuracy within wireless communication systems.
Authors: Alakananda Mitra, Sahila Beegum, David Fleisher, Vangimalla R. Reddy, Wenguang Sun, Chittaranjan Ray, Dennis Timlin, Arindam Malakar
The cotton industry in the United States is committed to sustainable production practices that minimize water, land, and energy use while improving soil health and cotton output. Climate-smart agricultural technologies are being developed to boost yields while decreasing operating expenses. Crop yield prediction, on the other hand, is difficult because of the complex and nonlinear impacts of cultivar, soil type, management, pest and disease, climate, and weather patterns on crops. To solve this issue, we employ machine learning (ML) to forecast production while considering climate change, soil diversity, cultivar, and inorganic nitrogen levels. From the 1980s to the 1990s, field data were gathered across the southern cotton belt of the United States. To capture the most current effects of climate change over the previous six years, a second data source was produced using the process-based crop model, GOSSYM. We concentrated our efforts on three distinct areas inside each of the three southern states: Texas, Mississippi, and Georgia. To simplify the amount of computations, accumulated heat units (AHU) for each set of experimental data were employed as an analogy to use time-series weather data. The Random Forest Regressor yielded a 97.75% accuracy rate, with a root mean square error of 55.05 kg/ha and an R2 of around 0.98. These findings demonstrate how an ML technique may be developed and applied as a reliable and easy-to-use model to support the cotton climate-smart initiative.
Authors: Cheng Ding, Zhicheng Guo, Cynthia Rudin, Ran Xiao, Fadi B Nahab, Xiao Hu
This paper explores the challenges in evaluating machine learning (ML) models for continuous health monitoring using wearable devices beyond conventional metrics. We state the complexities posed by real-world variability, disease dynamics, user-specific characteristics, and the prevalence of false notifications, necessitating novel evaluation strategies. Drawing insights from large-scale heart studies, the paper offers a comprehensive guideline for robust ML model evaluation on continuous health monitoring.
Authors: Romain Lacombe, Lucas Hendren, Khalid El-Awady
A central challenge of the clean energy transition is the development of catalysts for low-emissions technologies. Recent advances in Machine Learning for quantum chemistry drastically accelerate the computation of catalytic activity descriptors such as adsorption energies. Here we introduce AdsorbRL, a Deep Reinforcement Learning agent aiming to identify potential catalysts given a multi-objective binding energy target, trained using offline learning on the Open Catalyst 2020 and Materials Project data sets. We experiment with Deep Q-Network agents to traverse the space of all ~160,000 possible unary, binary and ternary compounds of 55 chemical elements, with very sparse rewards based on adsorption energy known for only between 2,000 and 3,000 catalysts per adsorbate. To constrain the actions space, we introduce Random Edge Traversal and train a single-objective DQN agent on the known states subgraph, which we find strengthens target binding energy by an average of 4.1 eV. We extend this approach to multi-objective, goal-conditioned learning, and train a DQN agent to identify materials with the highest (respectively lowest) adsorption energies for multiple simultaneous target adsorbates. We experiment with Objective Sub-Sampling, a novel training scheme aimed at encouraging exploration in the multi-objective setup, and demonstrate simultaneous adsorption energy improvement across all target adsorbates, by an average of 0.8 eV. Overall, our results suggest strong potential for Deep Reinforcement Learning applied to the inverse catalysts design problem.
Authors: Sidney Tio, Jimmy Ho, Pradeep Varakantham
We adapt Parameterized Environment Response Model (PERM), a method for training both Reinforcement Learning (RL) Agents and human learners in parameterized environments by directly modeling difficulty and ability. Inspired by Item Response Theory (IRT), PERM aligns environment difficulty with individual ability, creating a Zone of Proximal Development-based curriculum. Remarkably, PERM operates without real-time RL updates and allows for offline training, ensuring its adaptability across diverse students. We present a two-stage training process that capitalizes on PERM's adaptability, and demonstrate its effectiveness in training RL agents and humans in an empirical study.
Authors: Yizhou Wang, Ruiyi Zhang, Haoliang Wang, Uttaran Bhattacharya, Yun Fu, Gang Wu
Recent advancements in language-model-based video understanding have been progressing at a remarkable pace, spurred by the introduction of Large Language Models (LLMs). However, the focus of prior research has been predominantly on devising a projection layer that maps video features to tokens, an approach that is both rudimentary and inefficient. In our study, we introduce a cutting-edge framework, VaQuitA, designed to refine the synergy between video and textual information. At the data level, instead of sampling frames uniformly, we implement a sampling method guided by CLIP-score rankings, which enables a more aligned selection of frames with the given question. At the feature level, we integrate a trainable Video Perceiver alongside a Visual-Query Transformer (abbreviated as VQ-Former), which bolsters the interplay between the input question and the video features. We also discover that incorporating a simple prompt, "Please be critical", into the LLM input can substantially enhance its video comprehension capabilities. Our experimental results indicate that VaQuitA consistently sets a new benchmark for zero-shot video question-answering tasks and is adept at producing high-quality, multi-turn video dialogues with users.
Authors: Lukas Schäfer, Logan Jones, Anssi Kanervisto, Yuhan Cao, Tabish Rashid, Raluca Georgescu, Dave Bignell, Siddhartha Sen, Andrea Treviño Gavito, Sam Devlin
Video games have served as useful benchmarks for the decision making community, but going beyond Atari games towards training agents in modern games has been prohibitively expensive for the vast majority of the research community. Recent progress in the research, development and open release of large vision models has the potential to amortize some of these costs across the community. However, it is currently unclear which of these models have learnt representations that retain information critical for sequential decision making. Towards enabling wider participation in the research of gameplaying agents in modern games, we present a systematic study of imitation learning with publicly available visual encoders compared to the typical, task-specific, end-to-end training approach in Minecraft, Minecraft Dungeons and Counter-Strike: Global Offensive.
Authors: Tong Xia, Abhirup Ghosh, Cecilia Mascolo
Learning a global model by abstracting the knowledge, distributed across multiple clients, without aggregating the raw data is the primary goal of Federated Learning (FL). Typically, this works in rounds alternating between parallel local training at several clients, followed by model aggregation at a server. We found that existing FL methods under-perform when local datasets are small and present severe label skew as these lead to over-fitting and local model bias. This is a realistic setting in many real-world applications. To address the problem, we propose \textit{FLea}, a unified framework that tackles over-fitting and local bias by encouraging clients to exchange privacy-protected features to aid local training. The features refer to activations from an intermediate layer of the model, which are obfuscated before being shared with other clients to protect sensitive information in the data. \textit{FLea} leverages a novel way of combining local and shared features as augmentations to enhance local model learning. Our extensive experiments demonstrate that \textit{FLea} outperforms the start-of-the-art FL methods, sharing only model parameters, by up to $17.6\%$, and FL methods that share data augmentations by up to $6.3\%$, while reducing the privacy vulnerability associated with shared data augmentations.
Authors: Carolina Zheng, Keyon Vafa, David M. Blei
A recent line of work in natural language processing has aimed to combine language models and topic models. These topic-guided language models augment neural language models with topic models, unsupervised learning methods that can discover document-level patterns of word use. This paper compares the effectiveness of these methods in a standardized setting. We study four topic-guided language models and two baselines, evaluating the held-out predictive performance of each model on four corpora. Surprisingly, we find that none of these methods outperform a standard LSTM language model baseline, and most fail to learn good topics. Further, we train a probe of the neural language model that shows that the baseline's hidden states already encode topic information. We make public all code used for this study.
Authors: Derek Lim, Joshua Robinson, Stefanie Jegelka, Haggai Maron
Recent work has shown the utility of developing machine learning models that respect the structure and symmetries of eigenvectors. These works promote sign invariance, since for any eigenvector v the negation -v is also an eigenvector. However, we show that sign invariance is theoretically limited for tasks such as building orthogonally equivariant models and learning node positional encodings for link prediction in graphs. In this work, we demonstrate the benefits of sign equivariance for these tasks. To obtain these benefits, we develop novel sign equivariant neural network architectures. Our models are based on a new analytic characterization of sign equivariant polynomials and thus inherit provable expressiveness properties. Controlled synthetic experiments show that our networks can achieve the theoretically predicted benefits of sign equivariant models. Code is available at https://github.com/cptq/Sign-Equivariant-Nets.
Authors: Anton Pollak, Rajesh Menon
Image-to-Image translation (I2I) is a subtype of Machine Learning (ML) that has tremendous potential in applications where two domains of images and the need for translation between the two exist, such as the removal of fog. For example, this could be useful for autonomous vehicles, which currently struggle with adverse weather conditions like fog. However, datasets for I2I tasks are not abundant and typically hard to acquire. Here, we introduce STEREOFOG, a dataset comprised of $10,067$ paired fogged and clear images, captured using a custom-built device, with the purpose of exploring I2I's potential in this domain. It is the only real-world dataset of this kind to the best of our knowledge. Furthermore, we apply and optimize the pix2pix I2I ML framework to this dataset. With the final model achieving an average Complex Wavelet-Structural Similarity (CW-SSIM) score of $0.76$, we prove the technique's suitability for the problem.
Authors: Oliver Limoyo, Abhisek Konar, Trevor Ablett, Jonathan Kelly, Francois R. Hogan, Gregory Dudek
We present Learning to Place by Picking (LPP), a method capable of autonomously collecting demonstrations for a family of placing tasks in which objects must be manipulated to specific locations. With LPP, we approach the learning of robotic object placement policies by reversing the grasping process and exploiting the inherent symmetry of the pick and place problems. Specifically, we obtain placing demonstrations from a set of grasp sequences of objects that are initially located at their target placement locations. Our system is capable of collecting hundreds of demonstrations without human intervention by using a combination of tactile sensing and compliant control for grasps. We train a policy directly from visual observations through behaviour cloning, using the autonomously-collected demonstrations. By doing so, the policy can generalize to object placement scenarios outside of the training environment without privileged information (e.g., placing a plate picked up from a table and not at the original placement location). We validate our approach on home robotic scenarios that include dishwasher loading and table setting. Our approach yields robotic placing policies that outperform policies trained with kinesthetic teaching, both in terms of performance and data efficiency, while requiring no human supervision.
Authors: Vincent Liu, Prabhat Nagarajan, Andrew Patterson, Martha White
Offline reinforcement learning algorithms often require careful hyperparameter tuning. Consequently, before deployment, we need to select amongst a set of candidate policies. As yet, however, there is little understanding about the fundamental limits of this offline policy selection (OPS) problem. In this work we aim to provide clarity on when sample efficient OPS is possible, primarily by connecting OPS to off-policy policy evaluation (OPE) and Bellman error (BE) estimation. We first show a hardness result, that in the worst case, OPS is just as hard as OPE, by proving a reduction of OPE to OPS. As a result, no OPS method can be more sample efficient than OPE in the worst case. We then propose a BE method for OPS, called Identifiable BE Selection (IBES), that has a straightforward method for selecting its own hyperparameters. We highlight that using IBES for OPS generally has more requirements than OPE methods, but if satisfied, can be more sample efficient. We conclude with an empirical study comparing OPE and IBES, and by showing the difficulty of OPS on an offline Atari benchmark dataset.
Authors: Lennart Brocki, Neo Christopher Chung
Interpretability methods are critical components for examining and exploring deep neural networks (DNN), as well as increasing our understanding of and trust in them. Vision transformers (ViT), which can be trained to state-of-the-art performance with a self-supervised learning (SSL) training method, provide built-in attention maps (AM). While AMs can provide high-quality semantic segmentation of input images, they do not account for any signal coming from a downstream classifier. We introduce class-discriminative attention maps (CDAM), a novel post-hoc explanation method that is highly sensitive to the target class. Our method essentially scales attention scores by how relevant the corresponding tokens are for the predictions of a classifier head. Alternative to classifier outputs, CDAM can also explain a user-defined concept by targeting similarity measures in the latent space of the ViT. This allows for explanations of arbitrary concepts, defined by the user through a few sample images. We investigate the operating characteristics of CDAM in comparison with relevance propagation (RP) and token ablation maps (TAM), an alternative to pixel occlusion methods. CDAM is highly class-discriminative and semantically relevant, while providing implicit regularization of relevance scores.
PyTorch implementation: \url{https://github.com/lenbrocki/CDAM}
Web live demo: \url{https://cdam.informatism.com/}
Authors: Tianle Zhong, Jiechen Zhao, Xindi Guo, Qiang Su, Geoffrey Fox
Deep learning datasets are expanding at an unprecedented pace, creating new challenges for data processing in model training pipelines. A crucial aspect of these pipelines is dataset shuffling, which significantly improves unbiased learning and convergence accuracy by adhering to the principles of random sampling. However, loading shuffled data for large datasets incurs significant overhead in the deep learning pipeline and severely impacts the end-to-end training throughput. To mitigate this, current deep learning systems often resort to partial dataset shuffling, sacrificing global randomness to maintain acceptable training throughput on large datasets, still leaving global shuffling efficiency issues not fully explored.
In this work, we present RINAS, a data loading framework that systematically addresses the performance bottleneck of loading global shuffled datasets. Our key contribution is to offer an intra-batch unordered data fetching approach, which unleashes unexplored parallelism of data loading. We implement RINAS under the PyTorch framework for common dataset libraries HuggingFace and TorchVision. Our experimental results show that RINAS improves the throughput of general language model training and vision model training by up to 59% and 89%, respectively.
Authors: Zhan Gao, Amanda Prorok, Elvin Isufi
Analyzing the stability of graph neural networks (GNNs) under topological perturbations is key to understanding their transferability and the role of each architecture component. However, stability has been investigated only for particular architectures, questioning whether it holds for a broader spectrum of GNNs or only for a few instances. To answer this question, we study the stability of EdgeNet: a general GNN framework that unifies more than twenty solutions including the convolutional and attention-based classes, as well as graph isomorphism networks and hybrid architectures. We prove that all GNNs within the EdgeNet framework are stable to topological perturbations. By studying the effect of different EdgeNet categories on the stability, we show that GNNs with fewer degrees of freedom in their parameter space, linked to a lower representational capacity, are more stable. The key factor yielding this trade-off is the eigenvector misalignment between the EdgeNet parameter matrices and the graph shift operator. For example, graph convolutional neural networks that assign a single scalar per signal shift (hence, with a perfect alignment) are more stable than the more involved node or edge-varying counterparts. Extensive numerical results corroborate our theoretical findings and highlight the role of different architecture components in the trade-off.
Authors: Ting-Yu Dai, Dev Niyogi, Zoltan Nagy
Urban Building Energy Modeling (UBEM) is an emerging method to investigate urban design and energy systems against the increasing energy demand at urban and neighborhood levels. However, current UBEM methods are mostly physic-based and time-consuming in multiple climate change scenarios. This work proposes CityTFT, a data-driven UBEM framework, to accurately model the energy demands in urban environments. With the empowerment of the underlying TFT framework and an augmented loss function, CityTFT could predict heating and cooling triggers in unseen climate dynamics with an F1 score of 99.98 \% while RMSE of loads of 13.57 kWh.
Authors: Anthony Zhou, Amir Barati Farimani
The growth of deep learning in the past decade has motivated important applications to smart manufacturing and machine health monitoring. In particular, vibration data offers a rich and reliable source to provide meaningful insights into machine health and predictive maintenance. In this work, we present a Transformer based framework for analyzing vibration signals to predict different types of bearing faults (FaultFormer). In particular, we process signal data using data augmentations and extract their Fourier modes to train a transformer encoder to achieve state of the art accuracies. The attention mechanism as well as model outputs were analyzed to confirm the transformer's ability to automatically extract features within signals and learn both global and local relationships to make classifications. Lastly, two pretraining strategies were proposed to pave the way for large, generalizable transformers that could adapt to new data, situations, or machinery on the production floor.
Authors: Regina de Brito Duarte, Qiwei Han, Claudia Soares
Medical referrals between primary care physicians (PC) and specialist care (SC) physicians profoundly impact patient care regarding quality, satisfaction, and cost. This paper investigates the influence of professional networks among medical doctors on referring patients from PC to SC. Using five-year consultation data from a Portuguese private health provider, we conducted exploratory data analysis and constructed both professional and referral networks among physicians. We then apply Graph Neural Network (GNN) models to learn latent representations of the referral network. Our analysis supports the hypothesis that doctors' professional social connections can predict medical referrals, potentially enhancing collaboration within organizations and improving healthcare services. This research contributes to dissecting the underlying mechanisms in primary-specialty referrals, thereby providing valuable insights for enhancing patient care and effective healthcare management.
Authors: Jamie Santos, Holly Dinkel, Julia Di, Paulo V.K. Borges, Marina Moreira, Oleg Alexandrov, Brian Coltin, Trey Smith
This work presents an algorithm for scene change detection from point clouds to enable autonomous robotic caretaking in future space habitats. Autonomous robotic systems will help maintain future deep-space habitats, such as the Gateway space station, which will be uncrewed for extended periods. Existing scene analysis software used on the International Space Station (ISS) relies on manually-labeled images for detecting changes. In contrast, the algorithm presented in this work uses raw, unlabeled point clouds as inputs. The algorithm first applies modified Expectation-Maximization Gaussian Mixture Model (GMM) clustering to two input point clouds. It then performs change detection by comparing the GMMs using the Earth Mover's Distance. The algorithm is validated quantitatively and qualitatively using a test dataset collected by an Astrobee robot in the NASA Ames Granite Lab comprising single frame depth images taken directly by Astrobee and full-scene reconstructed maps built with RGB-D and pose data from Astrobee. The runtimes of the approach are also analyzed in depth. The source code is publicly released to promote further development.
Authors: Sai Venkatesh Chilukoti, Md Imran Hossen, Liqun Shan, Vijay Srinivas Tida, Xiai Hei
DP-SGD has emerged as a popular method to protect personally identifiable information in deep learning applications. Unfortunately, DP-SGD's per-sample gradient clipping and uniform noise addition during training can significantly degrade model utility. To enhance the model's utility, researchers proposed various adaptive DP-SGD methods. However, we examine and discover that these techniques result in greater privacy leakage or lower accuracy than the traditional DP-SGD method, or a lack of evaluation on a complex data set such as CIFAR100. To address these limitations, we propose an Auto DP-SGD. Our method automates clipping threshold estimation based on the DL model's gradient norm and scales the gradients of each training sample without losing gradient information. This helps to improve the algorithm's utility while using a less privacy budget. To further improve accuracy, we introduce automatic noise multiplier decay mechanisms to decrease the noise multiplier after every epoch. Finally, we develop closed-form mathematical expressions using tCDP accountant for automatic noise multiplier and automatic clipping threshold estimation. Through extensive experimentation, we demonstrate that Auto DP-SGD outperforms existing SOTA DP-SGD methods in privacy and accuracy on various benchmark datasets. We also show that privacy can be improved by lowering the scale factor and using learning rate schedulers without significantly reducing accuracy. Specifically, Auto DP-SGD, when used with a step noise multiplier, improves accuracy by 3.20, 1.57, 6.73, and 1.42 for the MNIST, CIFAR10, CIFAR100, and AG News Corpus datasets, respectively. Furthermore, it obtains a substantial reduction in the privacy budget of 94.9, 79.16, 67.36, and 53.37 for the corresponding data sets.
Authors: Alex J. Chan, José Luis Redondo García, Fabrizio Silvestri, Colm O'Donnel, Konstantina Palla
Content moderation at scale faces the challenge of considering local cultural distinctions when assessing content. While global policies aim to maintain decision-making consistency and prevent arbitrary rule enforcement, they often overlook regional variations in interpreting natural language as expressed in content. In this study, we are looking into how moderation systems can tackle this issue by adapting to local comprehension nuances. We train large language models on extensive datasets of media news and articles to create culturally attuned models. The latter aim to capture the nuances of communication across geographies with the goal of recognizing cultural and societal variations in what is considered offensive content. We further explore the capability of these models to generate explanations for instances of content violation, aiming to shed light on how policy guidelines are perceived when cultural and societal contexts change. We find that training on extensive media datasets successfully induced cultural awareness and resulted in improvements in handling content violations on a regional basis. Additionally, these advancements include the ability to provide explanations that align with the specific local norms and nuances as evidenced by the annotators' preference in our conducted study. This multifaceted success reinforces the critical role of an adaptable content moderation approach in keeping pace with the ever-evolving nature of the content it oversees.
Authors: Alon Albalak, Liangming Pan, Colin Raffel, William Yang Wang
The data used to pretrain large language models has a decisive impact on a model's downstream performance, which has led to a large body of work on data selection methods that aim to automatically determine the most suitable data to use for pretraining. Existing data selection methods suffer from slow and computationally expensive processes, a problem amplified by the increasing size of models and of pretraining datasets. Data mixing, on the other hand, reduces the complexity of data selection by grouping data points together and determining sampling probabilities across entire groups. However, data mixing proportions are typically fixed before training and therefore cannot adapt to changing training dynamics. To address these limitations, we develop an efficient algorithm for Online Data Mixing (ODM) that combines elements from both data selection and data mixing. Based on multi-armed bandit algorithms, our online approach optimizes the data mixing proportions during training. Remarkably, our method trains a model that reaches the final perplexity of the next best method with 19\% fewer training iterations, and improves performance on the 5-shot MMLU benchmark by 1.9% relative accuracy, while adding negligible wall-clock time during pretraining.
Authors: Lulu Ge, Keshab K. Parhi
This paper addresses the clustering of data in the hyperdimensional computing (HDC) domain. In prior work, an HDC-based clustering framework, referred to as HDCluster, has been proposed. However, the performance of the existing HDCluster is not robust. The performance of HDCluster is degraded as the hypervectors for the clusters are chosen at random during the initialization step. To overcome this bottleneck, we assign the initial cluster hypervectors by exploring the similarity of the encoded data, referred to as \textit{query} hypervectors. Intra-cluster hypervectors have a higher similarity than inter-cluster hypervectors. Harnessing the similarity results among query hypervectors, this paper proposes four HDC-based clustering algorithms: similarity-based k-means, equal bin-width histogram, equal bin-height histogram, and similarity-based affinity propagation. Experimental results illustrate that: (i) Compared to the existing HDCluster, our proposed HDC-based clustering algorithms can achieve better accuracy, more robust performance, fewer iterations, and less execution time. Similarity-based affinity propagation outperforms the other three HDC-based clustering algorithms on eight datasets by 2~38% in clustering accuracy. (ii) Even for one-pass clustering, i.e., without any iterative update of the cluster hypervectors, our proposed algorithms can provide more robust clustering accuracy than HDCluster. (iii) Over eight datasets, five out of eight can achieve higher or comparable accuracy when projected onto the hyperdimensional space. Traditional clustering is more desirable than HDC when the number of clusters, $k$, is large.
Authors: Jinqian Chen, Jihua Zhu, Qinghai Zheng
Federated learning encounters a critical challenge of data heterogeneity, adversely affecting the performance and convergence of the federated model. Various approaches have been proposed to address this issue, yet their effectiveness is still limited. Recent studies have revealed that the federated model suffers severe forgetting in local training, leading to global forgetting and performance degradation. Although the analysis provides valuable insights, a comprehensive understanding of the vulnerable classes and their impact factors is yet to be established. In this paper, we aim to bridge this gap by systematically analyzing the forgetting degree of each class during local training across different communication rounds. Our observations are: (1) Both missing and non-dominant classes suffer similar severe forgetting during local training, while dominant classes show improvement in performance. (2) When dynamically reducing the sample size of a dominant class, catastrophic forgetting occurs abruptly when the proportion of its samples is below a certain threshold, indicating that the local model struggles to leverage a few samples of a specific class effectively to prevent forgetting. Motivated by these findings, we propose a novel and straightforward algorithm called Federated Knowledge Anchor (FedKA). Assuming that all clients have a single shared sample for each class, the knowledge anchor is constructed before each local training stage by extracting shared samples for missing classes and randomly selecting one sample per class for non-dominant classes. The knowledge anchor is then utilized to correct the gradient of each mini-batch towards the direction of preserving the knowledge of the missing and non-dominant classes. Extensive experimental results demonstrate that our proposed FedKA achieves fast and stable convergence, significantly improving accuracy on popular benchmarks.
Authors: Yu Yang, Aaditya K. Singh, Mostafa Elhoushi, Anas Mahmoud, Kushal Tirumala, Fabian Gloeckle, Baptiste Rozière, Carole-Jean Wu, Ari S. Morcos, Newsha Ardalani
Code datasets, often collected from diverse and uncontrolled sources such as GitHub, potentially suffer from quality issues, thereby affecting the performance and training efficiency of Large Language Models (LLMs) optimized for code generation. Previous studies demonstrated the benefit of using embedding spaces for data pruning, but they mainly focused on duplicate removal or increasing variety, and in other modalities, such as images. Our work focuses on using embeddings to identify and remove "low-quality" code data. First, we explore features of "low-quality" code in embedding space, through the use of synthetic corruptions. Armed with this knowledge, we devise novel pruning metrics that operate in embedding space to identify and remove low-quality entries in the Stack dataset. We demonstrate the benefits of this synthetic corruption informed pruning (SCIP) approach on the well-established HumanEval and MBPP benchmarks, outperforming existing embedding-based methods. Importantly, we achieve up to a 3% performance improvement over no pruning, thereby showing the promise of insights from synthetic corruptions for data pruning.
Authors: Wei-Cheng Chang, Jyun-Yu Jiang, Jiong Zhang, Mutasem Al-Darabsah, Choon Hui Teo, Cho-Jui Hsieh, Hsiang-Fu Yu, S.V.N. Vishwanathan
Embedding-based Retrieval Models (ERMs) have emerged as a promising framework for large-scale text retrieval problems due to powerful large language models. Nevertheless, fine-tuning ERMs to reach state-of-the-art results can be expensive due to the extreme scale of data as well as the complexity of multi-stages pipelines (e.g., pre-training, fine-tuning, distillation). In this work, we propose the PEFA framework, namely ParamEter-Free Adapters, for fast tuning of ERMs without any backward pass in the optimization. At index building stage, PEFA equips the ERM with a non-parametric k-nearest neighbor (kNN) component. At inference stage, PEFA performs a convex combination of two scoring functions, one from the ERM and the other from the kNN. Based on the neighborhood definition, PEFA framework induces two realizations, namely PEFA-XL (i.e., extra large) using double ANN indices and PEFA-XS (i.e., extra small) using a single ANN index. Empirically, PEFA achieves significant improvement on two retrieval applications. For document retrieval, regarding Recall@100 metric, PEFA improves not only pre-trained ERMs on Trivia-QA by an average of 13.2%, but also fine-tuned ERMs on NQ-320K by an average of 5.5%, respectively. For product search, PEFA improves the Recall@100 of the fine-tuned ERMs by an average of 5.3% and 14.5%, for PEFA-XS and PEFA-XL, respectively. Our code is available at https://github.com/ amzn/pecos/tree/mainline/examples/pefa-wsdm24
Authors: Yash Chandak, Shiv Shankar, Vasilis Syrgkanis, Emma Brunskill
Indirect experiments provide a valuable framework for estimating treatment effects in situations where conducting randomized control trials (RCTs) is impractical or unethical. Unlike RCTs, indirect experiments estimate treatment effects by leveraging (conditional) instrumental variables, enabling estimation through encouragement and recommendation rather than strict treatment assignment. However, the sample efficiency of such estimators depends not only on the inherent variability in outcomes but also on the varying compliance levels of users with the instrumental variables and the choice of estimator being used, especially when dealing with numerous instrumental variables. While adaptive experiment design has a rich literature for direct experiments, in this paper we take the initial steps towards enhancing sample efficiency for indirect experiments by adaptively designing a data collection policy over instrumental variables. Our main contribution is a practical computational procedure that utilizes influence functions to search for an optimal data collection policy, minimizing the mean-squared error of the desired (non-linear) estimator. Through experiments conducted in various domains inspired by real-world applications, we showcase how our method can significantly improve the sample efficiency of indirect experiments.
Authors: Chao Wang, Alexandre Hoang Thiery
This article introduces GIT-Net, a deep neural network architecture for approximating Partial Differential Equation (PDE) operators, inspired by integral transform operators. GIT-NET harnesses the fact that differential operators commonly used for defining PDEs can often be represented parsimoniously when expressed in specialized functional bases (e.g., Fourier basis). Unlike rigid integral transforms, GIT-Net parametrizes adaptive generalized integral transforms with deep neural networks. When compared to several recently proposed alternatives, GIT-Net's computational and memory requirements scale gracefully with mesh discretizations, facilitating its application to PDE problems on complex geometries. Numerical experiments demonstrate that GIT-Net is a competitive neural network operator, exhibiting small test errors and low evaluations across a range of PDE problems. This stands in contrast to existing neural network operators, which typically excel in just one of these areas.
Authors: Weiming Xu, Tao Yang, Peng Zhang
Oscillatory combustion in aero engines and modern gas turbines often has significant adverse effects on their operation, and accurately recognizing various oscillation modes is the prerequisite for understanding and controlling combustion instability. However, the high-dimensional spatial-temporal data of a complex combustion system typically poses considerable challenges to the dynamical mode recognition. Based on a two-layer bidirectional long short-term memory variational autoencoder (Bi-LSTM-VAE) dimensionality reduction model and a two-dimensional Wasserstein distance-based classifier (WDC), this study proposes a promising method (Bi-LSTM-VAE-WDC) for recognizing dynamical modes in oscillatory combustion systems. Specifically, the Bi-LSTM-VAE dimension reduction model was introduced to reduce the high-dimensional spatial-temporal data of the combustion system to a low-dimensional phase space; Gaussian kernel density estimates (GKDE) were computed based on the distribution of phase points in a grid; two-dimensional WD values were calculated from the GKDE maps to recognize the oscillation modes. The time-series data used in this study were obtained from numerical simulations of circular arrays of laminar flame oscillators. The results show that the novel Bi-LSTM-VAE method can produce a non-overlapping distribution of phase points, indicating an effective unsupervised mode recognition and classification. Furthermore, the present method exhibits a more prominent performance than VAE and PCA (principal component analysis) for distinguishing dynamical modes in complex flame systems, implying its potential in studying turbulent combustion.
This paper studies the fundamental learning problem of the energy-based model (EBM). Learning the EBM can be achieved using the maximum likelihood estimation (MLE), which typically involves the Markov Chain Monte Carlo (MCMC) sampling, such as the Langevin dynamics. However, the noise-initialized Langevin dynamics can be challenging in practice and hard to mix. This motivates the exploration of joint training with the generator model where the generator model serves as a complementary model to bypass MCMC sampling. However, such a method can be less accurate than the MCMC and result in biased EBM learning. While the generator can also serve as an initializer model for better MCMC sampling, its learning can be biased since it only matches the EBM and has no access to empirical training examples. Such biased generator learning may limit the potential of learning the EBM. To address this issue, we present a joint learning framework that interweaves the maximum likelihood learning algorithm for both the EBM and the complementary generator model. In particular, the generator model is learned by MLE to match both the EBM and the empirical data distribution, making it a more informative initializer for MCMC sampling of EBM. Learning generator with observed examples typically requires inference of the generator posterior. To ensure accurate and efficient inference, we adopt the MCMC posterior sampling and introduce a complementary inference model to initialize such latent MCMC sampling. We show that three separate models can be seamlessly integrated into our joint framework through two (dual-) MCMC teaching, enabling effective and efficient EBM learning.
Authors: Runpeng Yu, Xinchao Wang
In this paper, we make a bold attempt toward an ambitious task: given a pre-trained classifier, we aim to reconstruct an image generator, without relying on any data samples. From a black-box perspective, this challenge seems intractable, since it inevitably involves identifying the inverse function for a classifier, which is, by nature, an information extraction process. As such, we resort to leveraging the knowledge encapsulated within the parameters of the neural network. Grounded on the theory of Maximum-Margin Bias of gradient descent, we propose a novel learning paradigm, in which the generator is trained to ensure that the convergence conditions of the network parameters are satisfied over the generated distribution of the samples. Empirical validation from various image generation tasks substantiates the efficacy of our strategy.
Authors: Zhongyuan Zhao, Jake Perazzone, Gunjan Verma, Santiago Segarra
Computational offloading has become an enabling component for edge intelligence in mobile and smart devices. Existing offloading schemes mainly focus on mobile devices and servers, while ignoring the potential network congestion caused by tasks from multiple mobile devices, especially in wireless multi-hop networks. To fill this gap, we propose a low-overhead, congestion-aware distributed task offloading scheme by augmenting a distributed greedy framework with graph-based machine learning. In simulated wireless multi-hop networks with 20-110 nodes and a resource allocation scheme based on shortest path routing and contention-based link scheduling, our approach is demonstrated to be effective in reducing congestion or unstable queues under the context-agnostic baseline, while improving the execution latency over local computing.
Authors: Chaoyi Chen, Dechao Gao, Yanfeng Zhang, Qiange Wang, Zhenbo Fu, Xuecang Zhang, Junhua Zhu, Yu Gu, Ge Yu
Existing Graph Neural Network (GNN) training frameworks have been designed to help developers easily create performant GNN implementations. However, most existing GNN frameworks assume that the input graphs are static, but ignore that most real-world graphs are constantly evolving. Though many dynamic GNN models have emerged to learn from evolving graphs, the training process of these dynamic GNNs is dramatically different from traditional GNNs in that it captures both the spatial and temporal dependencies of graph updates. This poses new challenges for designing dynamic GNN training frameworks. First, the traditional batched training method fails to capture real-time structural evolution information. Second, the time-dependent nature makes parallel training hard to design. Third, it lacks system supports for users to efficiently implement dynamic GNNs. In this paper, we present NeutronStream, a framework for training dynamic GNN models. NeutronStream abstracts the input dynamic graph into a chronologically updated stream of events and processes the stream with an optimized sliding window to incrementally capture the spatial-temporal dependencies of events. Furthermore, NeutronStream provides a parallel execution engine to tackle the sequential event processing challenge to achieve high performance. NeutronStream also integrates a built-in graph storage structure that supports dynamic updates and provides a set of easy-to-use APIs that allow users to express their dynamic GNNs. Our experimental results demonstrate that, compared to state-of-the-art dynamic GNN implementations, NeutronStream achieves speedups ranging from 1.48X to 5.87X and an average accuracy improvement of 3.97%.
Authors: Nesrine Cherif, Wael Jaafar, Halim Yanikomeroglu, Abbas Yongacoglu
Unmanned aerial vehicle (UAV) is a promising technology for last-mile cargo delivery. However, the limited on-board battery capacity, cellular unreliability, and frequent handoffs in the airspace are the main obstacles to unleash its full potential. Given that existing cellular networks were primarily designed to service ground users, re-utilizing the same architecture for highly mobile aerial users, e.g., cargo-UAVs, is deemed challenging. Indeed, to ensure a safe delivery using cargo-UAVs, it is crucial to utilize the available energy efficiently, while guaranteeing reliable connectivity for command-and-control and avoiding frequent handoff. To achieve this goal, we propose a novel approach for joint cargo-UAV trajectory planning and cell association. Specifically, we formulate the cargo-UAV mission as a multi-objective problem aiming to 1) minimize energy consumption, 2) reduce handoff events, and 3) guarantee cellular reliability along the trajectory. We leverage reinforcement learning (RL) to jointly optimize the cargo-UAV's trajectory and cell association. Simulation results demonstrate a performance improvement of our proposed method, in terms of handoffs, disconnectivity, and energy consumption, compared to benchmarks.
Authors: Phai Vu Dinh, Quang Uy Nguyen, Dinh Thai Hoang, Diep N. Nguyen, Son Pham Bao, Eryk Dutkiewicz
Intrusion detection systems (IDSs) play a critical role in protecting billions of IoT devices from malicious attacks. However, the IDSs for IoT devices face inherent challenges of IoT systems, including the heterogeneity of IoT data/devices, the high dimensionality of training data, and the imbalanced data. Moreover, the deployment of IDSs on IoT systems is challenging, and sometimes impossible, due to the limited resources such as memory/storage and computing capability of typical IoT devices. To tackle these challenges, this article proposes a novel deep neural network/architecture called Constrained Twin Variational Auto-Encoder (CTVAE) that can feed classifiers of IDSs with more separable/distinguishable and lower-dimensional representation data. Additionally, in comparison to the state-of-the-art neural networks used in IDSs, CTVAE requires less memory/storage and computing power, hence making it more suitable for IoT IDS systems. Extensive experiments with the 11 most popular IoT botnet datasets show that CTVAE can boost around 1% in terms of accuracy and Fscore in detection attack compared to the state-of-the-art machine learning and representation learning methods, whilst the running time for attack detection is lower than 2E-6 seconds and the model size is lower than 1 MB. We also further investigate various characteristics of CTVAE in the latent space and in the reconstruction representation to demonstrate its efficacy compared with current well-known methods.
Authors: Zhangyue Shi, Tianxin Xie, Chenang Liu, Yuxuan Li
The incorporation of advanced sensors and machine learning techniques has enabled modern manufacturing enterprises to perform data-driven in-situ quality monitoring based on the sensor data collected in manufacturing processes. However, one critical challenge is that newly presented defect category may manifest as the manufacturing process continues, resulting in monitoring performance deterioration of previously trained machine learning models. Hence, there is an increasing need for empowering machine learning model to learn continually. Among all continual learning methods, memory-based continual learning has the best performance but faces the constraints of data storage capacity. To address this issue, this paper develops a novel pseudo replay-based continual learning by integrating class incremental learning and oversampling-based data generation. Without storing all the data, the developed framework could generate high-quality data representing previous classes to train machine learning model incrementally when new category anomaly occurs. In addition, it could even enhance the monitoring performance since it also effectively improves the data quality. The effectiveness of the proposed framework is validated in an additive manufacturing process, which leverages supervised classification problem for anomaly detection. The experimental results show that the developed method is very promising in detecting novel anomaly while maintaining a good performance on the previous task and brings up more flexibility in model architecture.
Authors: Fumio Hashimoto, Kibo Ote
[Objective] This study aims to introduce a novel back projection-induced U-Net-shaped architecture, called ReconU-Net, for deep learning-based direct positron emission tomography (PET) image reconstruction. Additionally, our objective is to analyze the behavior of direct PET image reconstruction and gain deeper insights by comparing the proposed ReconU-Net architecture with other encoder-decoder architectures without skip connections. [Approach] The proposed ReconU-Net architecture uniquely integrates the physical model of the back projection operation into the skip connection. This distinctive feature facilitates the effective transfer of intrinsic spatial information from the input sinogram to the reconstructed image via an embedded physical model. The proposed ReconU-Net was trained using Monte Carlo simulation data from the Brainweb phantom and tested on both simulated and real Hoffman brain phantom data. [Main results] The proposed ReconU-Net method generated a reconstructed image with a more accurate structure compared to other deep learning-based direct reconstruction methods. Further analysis showed that the proposed ReconU-Net architecture has the ability to transfer features of multiple resolutions, especially non-abstract high-resolution information, through skip connections. Despite limited training on simulated data, the proposed ReconU-Net successfully reconstructed the real Hoffman brain phantom, unlike other deep learning-based direct reconstruction methods, which failed to produce a reconstructed image. [Significance] The proposed ReconU-Net can improve the fidelity of direct PET image reconstruction, even when dealing with small training datasets, by leveraging the synergistic relationship between data-driven modeling and the physics model of the imaging process.
Authors: Zhengmao Ye, Dengchun Li, Jingqi Tian, Tingfeng Lan, Jie Zuo, Lei Duan, Hui Lu, Yexi Jiang, Jian Sha, Ke Zhang, Mingjie Tang
Transformer-based large language models (LLMs) have demonstrated outstanding performance across diverse domains, particularly when fine-turned for specific domains. Recent studies suggest that the resources required for fine-tuning LLMs can be economized through parameter-efficient methods such as Low-Rank Adaptation (LoRA). While LoRA effectively reduces computational burdens and resource demands, it currently supports only a single-job fine-tuning setup.
In this paper, we present ASPEN, a high-throughput framework for fine-tuning LLMs. ASPEN efficiently trains multiple jobs on a single GPU using the LoRA method, leveraging shared pre-trained model and adaptive scheduling. ASPEN is compatible with transformer-based language models like LLaMA and ChatGLM, etc. Experiments show that ASPEN saves 53% of GPU memory when training multiple LLaMA-7B models on NVIDIA A100 80GB GPU and boosts training throughput by about 17% compared to existing methods when training with various pre-trained models on different GPUs. The adaptive scheduling algorithm reduces turnaround time by 24%, end-to-end training latency by 12%, prioritizing jobs and preventing out-of-memory issues.
Authors: Ravid Shwartz-Ziv, Micah Goldblum, Yucen Lily Li, C. Bayan Bruss, Andrew Gordon Wilson
Real-world datasets are often highly class-imbalanced, which can adversely impact the performance of deep learning models. The majority of research on training neural networks under class imbalance has focused on specialized loss functions, sampling techniques, or two-stage training procedures. Notably, we demonstrate that simply tuning existing components of standard deep learning pipelines, such as the batch size, data augmentation, optimizer, and label smoothing, can achieve state-of-the-art performance without any such specialized class imbalance methods. We also provide key prescriptions and considerations for training under class imbalance, and an understanding of why imbalance methods succeed or fail.
Authors: Chi Zhang, Penglin Cai, Yuhui Fu, Haoqi Yuan, Zongqing Lu
We study building embodied agents for open-ended creative tasks. While existing methods build instruction-following agents that can perform diverse open-ended tasks, none of them demonstrates creativity -- the ability to give novel and diverse task solutions implicit in the language instructions. This limitation comes from their inability to convert abstract language instructions into concrete task goals in the environment and perform long-horizon planning for such complicated goals. Given the observation that humans perform creative tasks with the help of imagination, we propose a class of solutions for creative agents, where the controller is enhanced with an imaginator that generates detailed imaginations of task outcomes conditioned on language instructions. We introduce several approaches to implementing the components of creative agents. We implement the imaginator with either a large language model for textual imagination or a diffusion model for visual imagination. The controller can either be a behavior-cloning policy learned from data or a pre-trained foundation model generating executable codes in the environment. We benchmark creative tasks with the challenging open-world game Minecraft, where the agents are asked to create diverse buildings given free-form language instructions. In addition, we propose novel evaluation metrics for open-ended creative tasks utilizing GPT-4V, which holds many advantages over existing metrics. We perform a detailed experimental analysis of creative agents, showing that creative agents are the first AI agents accomplishing diverse building creation in the survival mode of Minecraft. Our benchmark and models are open-source for future research on creative agents (https://github.com/PKU-RL/Creative-Agents).
Authors: Xinyi Yang, Xinting Yang, Chao Yu, Jiayu Chen, Huazhong Yang, Yu Wang
We investigate the problem of decentralized multi-agent navigation tasks, where multiple agents need to reach initially unassigned targets in a limited time. Classical planning-based methods suffer from expensive computation overhead at each step and offer limited expressiveness for complex cooperation strategies. In contrast, reinforcement learning (RL) has recently become a popular paradigm for addressing this issue. However, RL struggles with low data efficiency and cooperation when directly exploring (nearly) optimal policies in the large search space, especially with an increased agent number (e.g., 10+ agents) or in complex environments (e.g., 3D simulators). In this paper, we propose Multi-Agent Scalable GNN-based P lanner (MASP), a goal-conditioned hierarchical planner for navigation tasks with a substantial number of agents. MASP adopts a hierarchical framework to divide a large search space into multiple smaller spaces, thereby reducing the space complexity and accelerating training convergence. We also leverage graph neural networks (GNN) to model the interaction between agents and goals, improving goal achievement. Besides, to enhance generalization capabilities in scenarios with unseen team sizes, we divide agents into multiple groups, each with a previously trained number of agents. The results demonstrate that MASP outperforms classical planning-based competitors and RL baselines, achieving a nearly 100% success rate with minimal training data in both multi-agent particle environments (MPE) with 50 agents and a quadrotor 3-dimensional environment (OmniDrones) with 20 agents. Furthermore, the learned policy showcases zero-shot generalization across unseen team sizes.
Authors: Junho Song, Keonwoo Kim, Jeonglyul Oh, Sungzoon Cho
Detecting anomalies in real-world multivariate time series data is challenging due to complex temporal dependencies and inter-variable correlations. Recently, reconstruction-based deep models have been widely used to solve the problem. However, these methods still suffer from an over-generalization issue and fail to deliver consistently high performance. To address this issue, we propose the MEMTO, a memory-guided Transformer using a reconstruction-based approach. It is designed to incorporate a novel memory module that can learn the degree to which each memory item should be updated in response to the input data. To stabilize the training procedure, we use a two-phase training paradigm which involves using K-means clustering for initializing memory items. Additionally, we introduce a bi-dimensional deviation-based detection criterion that calculates anomaly scores considering both input space and latent space. We evaluate our proposed method on five real-world datasets from diverse domains, and it achieves an average anomaly detection F1-score of 95.74%, significantly outperforming the previous state-of-the-art methods. We also conduct extensive experiments to empirically validate the effectiveness of our proposed model's key components.
Authors: Shun Kotoku, Takatomo Mihana, André Röhm, Ryoichi Horisaki, Makoto Naruse
Photonic accelerators have recently attracted soaring interest, harnessing the ultimate nature of light for information processing. Collective decision-making with a laser network, employing the chaotic and synchronous dynamics of optically interconnected lasers to address the competitive multi-armed bandit (CMAB) problem, is a highly compelling approach due to its scalability and experimental feasibility. We investigated essential network structures for collective decision-making through quantitative stability analysis. Moreover, we demonstrated the asymmetric preferences of players in the CMAB problem, extending its functionality to more practical applications. Our study highlights the capability and significance of machine learning built upon chaotic lasers and photonic devices.
Authors: Yoshiki Sato, Makiko Konoshima, Hirotaka Tamura, Jun Ohkubo
Ising formulations are widely utilized to solve combinatorial optimization problems, and a variety of quantum or semiconductor-based hardware has recently been made available. In combinatorial optimization problems, the existence of local minima in energy landscapes is problematic to use to seek the global minimum. We note that the aim of the optimization is not to obtain exact samplings from the Boltzmann distribution, and there is thus no need to satisfy detailed balance conditions. In light of this fact, we develop an algorithm to get out of the local minima efficiently while it does not yield the exact samplings. For this purpose, we utilize a feature that characterizes locality in the current state, which is easy to obtain with a type of specialized hardware. Furthermore, as the proposed algorithm is based on a rejection-free algorithm, the computational cost is low. In this work, after presenting the details of the proposed algorithm, we report the results of numerical experiments that demonstrate the effectiveness of the proposed feature and algorithm.
Authors: Yongho Shin, Changyeol Lee, Hyung-Chan An
The learning-augmented multi-option ski rental problem generalizes the classical ski rental problem in two ways: the algorithm is provided with a prediction on the number of days we can ski, and the ski rental options now come with a variety of rental periods and prices to choose from, unlike the classical two-option setting. Subsequent to the initial study of the multi-option ski rental problem (without learning augmentation) due to Zhang, Poon, and Xu, significant progress has been made for this problem recently in particular. The problem is very well understood when we relinquish one of the two generalizations -- for the learning-augmented classical ski rental problem, algorithms giving best-possible trade-off between consistency and robustness exist; for the multi-option ski rental problem without learning augmentation, deterministic/randomized algorithms giving the best-possible competitiveness have been found. However, in presence of both generalizations, there remained a huge gap between the algorithmic and impossibility results. In fact, for randomized algorithms, we did not have any nontrivial lower bounds on the consistency-robustness trade-off before.
This paper bridges this gap for both deterministic and randomized algorithms. For deterministic algorithms, we present a best-possible algorithm that completely matches the known lower bound. For randomized algorithms, we show the first nontrivial lower bound on the consistency-robustness trade-off, and also present an improved randomized algorithm. Our algorithm matches our lower bound on robustness within a factor of e/2 when the consistency is at most 1.086.
Authors: Tianchi Cai, Xierui Song, Jiyan Jiang, Fei Teng, Jinjie Gu, Guannan Zhang
Language model alignment is a cutting-edge technique in large language model training to align the model output to user's intent, e.g., being helpful and harmless. Recent alignment framework consists of two steps: supervised fine-tuning with demonstration data and preference learning with human preference data. Previous preference learning methods, such as RLHF and DPO, mainly focus on pair-wise preference data. However, in many real-world scenarios where human feedbacks are intrinsically point-wise, these methods will suffer from information loss or even fail. To fill this gap, in this paper, we first develop a preference learning method called point-wise DPO to tackle point-wise preference data. Further revelation on the connection between supervised fine-tuning and point-wise preference learning enables us to develop a unified framework for both human demonstration and point-wise preference data, which sheds new light on the construction of preference dataset. Extensive experiments on point-wise datasets with binary or continuous labels demonstrate the superior performance and efficiency of our proposed methods. A new dataset with high-quality demonstration samples on harmlessness is constructed and made publicly available.
Authors: Michael Igorevich Ivanitskiy, Alex F. Spies, Tilman Räuker, Guillaume Corlouer, Chris Mathwin, Lucia Quirke, Can Rager, Rusheb Shah, Dan Valentine, Cecilia Diniz Behn, Katsumi Inoue, Samy Wu Fung
Transformer models underpin many recent advances in practical machine learning applications, yet understanding their internal behavior continues to elude researchers. Given the size and complexity of these models, forming a comprehensive picture of their inner workings remains a significant challenge. To this end, we set out to understand small transformer models in a more tractable setting: that of solving mazes. In this work, we focus on the abstractions formed by these models and find evidence for the consistent emergence of structured internal representations of maze topology and valid paths. We demonstrate this by showing that the residual stream of only a single token can be linearly decoded to faithfully reconstruct the entire maze. We also find that the learned embeddings of individual tokens have spatial structure. Furthermore, we take steps towards deciphering the circuity of path-following by identifying attention heads (dubbed $\textit{adjacency heads}$), which are implicated in finding valid subsequent tokens.
Authors: Junjie Gao, Xiangyu Zheng, DongDong Wang, Zhixiang Huang, Bangqi Zheng, Kai Yang
Uplift modeling refers to the set of machine learning techniques that a manager may use to estimate customer uplift, that is, the net effect of an action on some customer outcome. By identifying the subset of customers for whom a treatment will have the greatest effect, uplift models assist decision-makers in optimizing resource allocations and maximizing overall returns. Accurately estimating customer uplift poses practical challenges, as it requires assessing the difference between two mutually exclusive outcomes for each individual. In this paper, we propose two innovative adaptations of the well-established Gradient Boosting Decision Trees (GBDT) algorithm, which learn the causal effect in a sequential way and overcome the counter-factual nature. Both approaches innovate existing techniques in terms of ensemble learning method and learning objectives, respectively. Experiments on large-scale datasets demonstrate the usefulness of the proposed methods, which often yielding remarkable improvements over base models. To facilitate the application, we develop the UTBoost, an end-to-end tree boosting system specifically designed for uplift modeling. The package is open source and has been optimized for training speed to meet the needs of real industrial applications.
Authors: Alexandru Ţifrea, Preethi Lahoti, Ben Packer, Yoni Halpern, Ahmad Beirami, Flavien Prost
Post-processing mitigation techniques for group fairness generally adjust the decision threshold of a base model in order to improve fairness. Methods in this family exhibit several advantages that make them appealing in practice: post-processing requires no access to the model training pipeline, is agnostic to the base model architecture, and offers a reduced computation cost compared to in-processing. Despite these benefits, existing methods face other challenges that limit their applicability: they require knowledge of the sensitive attributes at inference time and are oftentimes outperformed by in-processing. In this paper, we propose a general framework to transform any in-processing method with a penalized objective into a post-processing procedure. The resulting method is specifically designed to overcome the aforementioned shortcomings of prior post-processing approaches. Furthermore, we show theoretically and through extensive experiments on real-world data that the resulting post-processing method matches or even surpasses the fairness-error trade-off offered by the in-processing counterpart.
Authors: Anuradha Kumari, M. Tanveer
In the realm of machine learning, the data may contain additional attributes, known as privileged information (PI). The main purpose of PI is to assist in the training of the model and then utilize the acquired knowledge to make predictions for unseen samples. Support vector regression (SVR) is an effective regression model, however, it has a low learning speed due to solving a convex quadratic problem (QP) subject to a pair of constraints. In contrast, twin support vector regression (TSVR) is more efficient than SVR as it solves two QPs each subject to one set of constraints. However, TSVR and its variants are trained only on regular features and do not use privileged features for training. To fill this gap, we introduce a fusion of TSVR with learning using privileged information (LUPI) and propose a novel approach called twin support vector regression with privileged information (TSVR+). The regularization terms in the proposed TSVR+ capture the essence of statistical learning theory and implement the structural risk minimization principle. We use the successive overrelaxation (SOR) technique to solve the optimization problem of the proposed TSVR+, which enhances the training efficiency. As far as our knowledge extends, the integration of the LUPI concept into twin variants of regression models is a novel advancement. The numerical experiments conducted on UCI, stock and time series data collectively demonstrate the superiority of the proposed model.
Authors: Florian Kofler, Hendrik Möller, Josef A. Buchner, Ezequiel de la Rosa, Ivan Ezhov, Marcel Rosier, Isra Mekki, Suprosanna Shit, Moritz Negwer, Rami Al-Maskari, Ali Ertürk, Shankeeth Vinayahalingam, Fabian Isensee, Sarthak Pati, Daniel Rueckert, Jan S. Kirschke, Stefan K. Ehrlich, Annika Reinke, Bjoern Menze, Benedikt Wiestler, Marie Piraud
This paper introduces panoptica, a versatile and performance-optimized package designed for computing instance-wise segmentation quality metrics from 2D and 3D segmentation maps. panoptica addresses the limitations of existing metrics and provides a modular framework that complements the original intersection over union-based panoptic quality with other metrics, such as the distance metric Average Symmetric Surface Distance. The package is open-source, implemented in Python, and accompanied by comprehensive documentation and tutorials. panoptica employs a three-step metrics computation process to cover diverse use cases. The efficacy of panoptica is demonstrated on various real-world biomedical datasets, where an instance-wise evaluation is instrumental for an accurate representation of the underlying clinical task. Overall, we envision panoptica as a valuable tool facilitating in-depth evaluation of segmentation methods.
Authors: Shashi Raj Pandey, Pierre Pinson, Petar Popovski
Data markets facilitate decentralized data exchange for applications such as prediction, learning, or inference. The design of these markets is challenged by varying privacy preferences as well as data similarity among data owners. Related works have often overlooked how data similarity impacts pricing and data value through statistical information leakage. We demonstrate that data similarity and privacy preferences are integral to market design and propose a query-response protocol using local differential privacy for a two-party data acquisition mechanism. In our regression data market model, we analyze strategic interactions between privacy-aware owners and the learner as a Stackelberg game over the asked price and privacy factor. Finally, we numerically evaluate how data similarity affects market participation and traded data value.
Authors: Xuan Long Do, Yiran Zhao, Hannah Brown, Yuxi Xie, James Xu Zhao, Nancy F. Chen, Kenji Kawaguchi, Michael Qizhe Xie, Junxian He
We propose a new method, Adversarial In-Context Learning (adv-ICL), to optimize prompt for in-context learning (ICL) by employing one LLM as a generator, another as a discriminator, and a third as a prompt modifier. As in traditional adversarial learning, adv-ICL is implemented as a two-player game between the generator and discriminator, where the generator tries to generate realistic enough output to fool the discriminator. In each round, given an input prefixed by task instructions and several exemplars, the generator produces an output. The discriminator is then tasked with classifying the generator input-output pair as model-generated or real data. Based on the discriminator loss, the prompt modifier proposes possible edits to the generator and discriminator prompts, and the edits that most improve the adversarial loss are selected. We show that adv-ICL results in significant improvements over state-of-the-art prompt optimization techniques for both open and closed-source models on 11 generation and classification tasks including summarization, arithmetic reasoning, machine translation, data-to-text generation, and the MMLU and big-bench hard benchmarks. In addition, because our method uses pre-trained models and updates only prompts rather than model parameters, it is computationally efficient, easy to extend to any LLM and task, and effective in low-resource settings.
Authors: Sungik Choi, Hankook Lee, Honglak Lee, Moontae Lee
Novelty detection is a fundamental task of machine learning which aims to detect abnormal ($\textit{i.e.}$ out-of-distribution (OOD)) samples. Since diffusion models have recently emerged as the de facto standard generative framework with surprising generation results, novelty detection via diffusion models has also gained much attention. Recent methods have mainly utilized the reconstruction property of in-distribution samples. However, they often suffer from detecting OOD samples that share similar background information to the in-distribution data. Based on our observation that diffusion models can \emph{project} any sample to an in-distribution sample with similar background information, we propose \emph{Projection Regret (PR)}, an efficient novelty detection method that mitigates the bias of non-semantic information. To be specific, PR computes the perceptual distance between the test image and its diffusion-based projection to detect abnormality. Since the perceptual distance often fails to capture semantic changes when the background information is dominant, we cancel out the background bias by comparing it against recursive projections. Extensive experiments demonstrate that PR outperforms the prior art of generative-model-based novelty detection methods by a significant margin.
Authors: Wangbin Sun, Jintang Li, Liang Chen, Bingzhe Wu, Yatao Bian, Zibin Zheng
Graph contrastive learning (GCL) has emerged as a representative paradigm in graph self-supervised learning, where negative samples are commonly regarded as the key to preventing model collapse and producing distinguishable representations. Recent studies have shown that GCL without negative samples can achieve state-of-the-art performance as well as scalability improvement, with bootstrapped graph latent (BGRL) as a prominent step forward. However, BGRL relies on a complex architecture to maintain the ability to scatter representations, and the underlying mechanisms enabling the success remain largely unexplored. In this paper, we introduce an instance-level decorrelation perspective to tackle the aforementioned issue and leverage it as a springboard to reveal the potential unnecessary model complexity within BGRL. Based on our findings, we present SGCL, a simple yet effective GCL framework that utilizes the outputs from two consecutive iterations as positive pairs, eliminating the negative samples. SGCL only requires a single graph augmentation and a single graph encoder without additional parameters. Extensive experiments conducted on various graph benchmarks demonstrate that SGCL can achieve competitive performance with fewer parameters, lower time and space costs, and significant convergence speedup.
Authors: Jiahang Li, Yakun Song, Xiang Song, David Paul Wipf
Graph Neural Networks (GNNs) have displayed considerable promise in graph representation learning across various applications. The core learning process requires the initialization of model weight matrices within each GNN layer, which is typically accomplished via classic initialization methods such as Xavier initialization. However, these methods were originally motivated to stabilize the variance of hidden embeddings and gradients across layers of Feedforward Neural Networks (FNNs) and Convolutional Neural Networks (CNNs) to avoid vanishing gradients and maintain steady information flow. In contrast, within the GNN context classical initializations disregard the impact of the input graph structure and message passing on variance. In this paper, we analyze the variance of forward and backward propagation across GNN layers and show that the variance instability of GNN initializations comes from the combined effect of the activation function, hidden dimension, graph structure and message passing. To better account for these influence factors, we propose a new initialization method for Variance Instability Reduction within GNN Optimization (Virgo), which naturally tends to equate forward and backward variances across successive layers. We conduct comprehensive experiments on 15 datasets to show that Virgo can lead to superior model performance and more stable variance at initialization on node classification, link prediction and graph classification tasks. Codes are in https://github.com/LspongebobJH/virgo_icml2023.
Authors: Xiaobei Zou, Luolin Xiong, Yang Tang, Jurgen Kurths
Spatio-temporal forecasting in various domains, like traffic prediction and weather forecasting, is a challenging endeavor, primarily due to the difficulties in modeling propagation dynamics and capturing high-dimensional interactions among nodes. Despite the significant strides made by graph-based networks in spatio-temporal forecasting, there remain two pivotal factors closely related to forecasting performance that need further consideration: time delays in propagation dynamics and multi-scale high-dimensional interactions. In this work, we present a Series-Aligned Multi-Scale Graph Learning (SAMSGL) framework, aiming to enhance forecasting performance. In order to handle time delays in spatial interactions, we propose a series-aligned graph convolution layer to facilitate the aggregation of non-delayed graph signals, thereby mitigating the influence of time delays for the improvement in accuracy. To understand global and local spatio-temporal interactions, we develop a spatio-temporal architecture via multi-scale graph learning, which encompasses two essential components: multi-scale graph structure learning and graph-fully connected (Graph-FC) blocks. The multi-scale graph structure learning includes a global graph structure to learn both delayed and non-delayed node embeddings, as well as a local one to learn node variations influenced by neighboring factors. The Graph-FC blocks synergistically fuse spatial and temporal information to boost prediction accuracy. To evaluate the performance of SAMSGL, we conduct experiments on meteorological and traffic forecasting datasets, which demonstrate its effectiveness and superiority.
Authors: Mohammad Safeea, Pedro Neto
This study evaluates the application of a discrete action space reinforcement learning method (Q-learning) to the continuous control problem of robot inverted pendulum balancing. To speed up the learning process and to overcome technical difficulties related to the direct learning on the real robotic system, the learning phase is performed in simulation environment. A mathematical model of the system dynamics is implemented, deduced by curve fitting on data acquired from the real system. The proposed approach demonstrated feasible, featuring its application on a real world robot that learned to balance an inverted pendulum. This study also reinforces and demonstrates the importance of an accurate representation of the physical world in simulation to achieve a more efficient implementation of reinforcement learning algorithms in real world, even when using a discrete action space algorithm to control a continuous action.
Authors: Foma Shipilov, Alexander Barnyakov, Vladimir Bobrovnikov, Sergey Kononov, Fedor Ratnikov
Particle identification at the Super Charm-Tau factory experiment will be provided by a Focusing Aerogel Ring Imaging CHerenkov detector (FARICH). The specifics of detector location make proper cooling difficult, therefore a significant number of ambient background hits are captured. They must be mitigated to reduce the data flow and improve particle velocity resolution. In this work we present several approaches to filtering signal hits, inspired by machine learning techniques from computer vision.
Authors: Andrew J. Charlton-Perez, Helen F. Dacre, Simon Driscoll, Suzanne L. Gray, Ben Harvey, Natalie J. Harvey, Kieran M. R. Hunt, Robert W. Lee, Ranjini Swaminathan, Remy Vandaele, Ambrogio Volonté
There has been huge recent interest in the potential of making operational weather forecasts using machine learning techniques. As they become a part of the weather forecasting toolbox, there is a pressing need to understand how well current machine learning models can simulate high-impactweather events. We compare forecasts of Storm Ciar\'an, a European windstorm that caused sixteen deaths and extensive damage in Northern Europe, made by machine learning and numericalweather prediction models. The four machine learning models considered (FourCastNet, Pangu-Weather, GraphCast and FourCastNet-v2) produce forecasts that accurately capture the synoptic-scale structure of the cyclone including the position of the cloud head, shape of the warm sector and location of warm conveyor belt jet, and the large-scale dynamical drivers important for the rapid storm development such as the position of the storm relative to the upper-level jet exit. However, their ability to resolve the more detailed structures important for issuing weather warnings is more mixed. All of the machine learning models underestimate the peak amplitude of winds associated with the storm, only some machine learning models resolve the warm core seclusion and none of the machine learning models capture the sharp bent-back warm frontal gradient. Our study shows there is a great deal about the performance and properties of machine learning weather forecasts that can be derived from case studies of high-impact weather events such as Storm Ciar\'an.
Authors: Pere Izquierdo Gomez, Miguel E. Lopez Gajardo, Nenad Mijatovic, Tomislav Dragicevic
Ensuring the reliability of power electronic converters is a matter of great importance, and data-driven condition monitoring techniques are cementing themselves as an important tool for this purpose. However, translating methods that work well in controlled lab environments to field applications presents significant challenges, notably because of the limited diversity and accuracy of the lab training data. By enabling the use of field data, online machine learning can be a powerful tool to overcome this problem, but it introduces additional challenges in ensuring the stability and predictability of the training processes. This work presents an edge computing method that mitigates these shortcomings with minimal additional memory usage, by employing an autonomous algorithm that prioritizes the storage of training samples with larger prediction errors. The method is demonstrated on the use case of a self-commissioning condition monitoring system, in the form of a thermal anomaly detection scheme for a variable frequency motor drive, where the algorithm self-learned to distinguish normal and anomalous operation with minimal prior knowledge. The obtained results, based on experimental data, show a significant improvement in prediction accuracy and training speed, when compared to equivalent models trained online without the proposed data selection process.
Authors: N. Ordonez, M. Tromp, P. M. Julbe, W. Böhmer
Agents trained with DQN rely on an observation at each timestep to decide what action to take next. However, in real world applications observations can change or be missing entirely. Examples of this could be a light bulb breaking down, or the wallpaper in a certain room changing. While these situations change the actual observation, the underlying optimal policy does not change. Because of this we want our agent to continue taking actions until it receives a (recognized) observation again. To achieve this we introduce a combination of a neural network architecture that uses hidden representations of the observations and a novel n-step loss function. Our implementation is able to withstand location based blindness stretches longer than the ones it was trained on, and therefore shows robustness to temporary blindness. For access to our implementation, please email Nathan, Marije, or Pau.
Authors: Tjeerd Jan Heeringa, Tim Roith, Christoph Brune, Martin Burger
This paper presents a method for finding a sparse representation of Barron functions. Specifically, given an $L^2$ function $f$, the inverse scale space flow is used to find a sparse measure $\mu$ minimising the $L^2$ loss between the Barron function associated to the measure $\mu$ and the function $f$. The convergence properties of this method are analysed in an ideal setting and in the cases of measurement noise and sampling bias. In an ideal setting the objective decreases strictly monotone in time to a minimizer with $\mathcal{O}(1/t)$, and in the case of measurement noise or sampling bias the optimum is achieved up to a multiplicative or additive constant. This convergence is preserved on discretization of the parameter space, and the minimizers on increasingly fine discretizations converge to the optimum on the full parameter space.
Authors: Mila Gorecki, Jakob H. Macke, Michael Deistler
Simulation-based inference (SBI) provides a powerful framework for inferring posterior distributions of stochastic simulators in a wide range of domains. In many settings, however, the posterior distribution is not the end goal itself -- rather, the derived parameter values and their uncertainties are used as a basis for deciding what actions to take. Unfortunately, because posterior distributions provided by SBI are (potentially crude) approximations of the true posterior, the resulting decisions can be suboptimal. Here, we address the question of how to perform Bayesian decision making on stochastic simulators, and how one can circumvent the need to compute an explicit approximation to the posterior. Our method trains a neural network on simulated data and can predict the expected cost given any data and action, and can, thus, be directly used to infer the action with lowest cost. We apply our method to several benchmark problems and demonstrate that it induces similar cost as the true posterior distribution. We then apply the method to infer optimal actions in a real-world simulator in the medical neurosciences, the Bayesian Virtual Epileptic Patient, and demonstrate that it allows to infer actions associated with low cost after few simulations.
Authors: Zhengyao Jiang, Yingchen Xu, Nolan Wagener, Yicheng Luo, Michael Janner, Edward Grefenstette, Tim Rocktäschel, Yuandong Tian
Humanoid control is an important research challenge offering avenues for integration into human-centric infrastructures and enabling physics-driven humanoid animations. The daunting challenges in this field stem from the difficulty of optimizing in high-dimensional action spaces and the instability introduced by the bipedal morphology of humanoids. However, the extensive collection of human motion-captured data and the derived datasets of humanoid trajectories, such as MoCapAct, paves the way to tackle these challenges. In this context, we present Humanoid Generalist Autoencoding Planner (H-GAP), a state-action trajectory generative model trained on humanoid trajectories derived from human motion-captured data, capable of adeptly handling downstream control tasks with Model Predictive Control (MPC). For 56 degrees of freedom humanoid, we empirically demonstrate that H-GAP learns to represent and generate a wide range of motor behaviours. Further, without any learning from online interactions, it can also flexibly transfer these behaviors to solve novel downstream control tasks via planning. Notably, H-GAP excels established MPC baselines that have access to the ground truth dynamics model, and is superior or comparable to offline RL methods trained for individual tasks. Finally, we do a series of empirical studies on the scaling properties of H-GAP, showing the potential for performance gains via additional data but not computing. Code and videos are available at https://ycxuyingchen.github.io/hgap/.
Authors: Philippe Gonzalez, Zheng-Hua Tan, Jan Østergaard, Jesper Jensen, Tommy Sonne Alstrøm, Tobias May
Diffusion models are a new class of generative models that have recently been applied to speech enhancement successfully. Previous works have demonstrated their superior performance in mismatched conditions compared to state-of-the art discriminative models. However, this was investigated with a single database for training and another one for testing, which makes the results highly dependent on the particular databases. Moreover, recent developments from the image generation literature remain largely unexplored for speech enhancement. These include several design aspects of diffusion models, such as the noise schedule or the reverse sampler. In this work, we systematically assess the generalization performance of a diffusion-based speech enhancement model by using multiple speech, noise and binaural room impulse response (BRIR) databases to simulate mismatched acoustic conditions. We also experiment with a noise schedule and a sampler that have not been applied to speech enhancement before. We show that the proposed system substantially benefits from using multiple databases for training, and achieves superior performance compared to state-of-the-art discriminative models in both matched and mismatched conditions. We also show that a Heun-based sampler achieves superior performance at a smaller computational cost compared to a sampler commonly used for speech enhancement.
Authors: Xiaze Zhang, Ziheng Ding, Qi Jing, Yuejie Zhang, Wenchao Ding, Rui Feng
Point clouds have shown significant potential in various domains, including Simultaneous Localization and Mapping (SLAM). However, existing approaches either rely on dense point clouds to achieve high localization accuracy or use generalized descriptors to reduce map size. Unfortunately, these two aspects seem to conflict with each other. To address this limitation, we propose a unified architecture, DeepPointMap, achieving excellent preference on both aspects. We utilize neural network to extract highly representative and sparse neural descriptors from point clouds, enabling memory-efficient map representation and accurate multi-scale localization tasks (e.g., odometry and loop-closure). Moreover, we showcase the versatility of our framework by extending it to more challenging multi-agent collaborative SLAM. The promising results obtained in these scenarios further emphasize the effectiveness and potential of our approach.
Authors: Tero Karras, Miika Aittala, Jaakko Lehtinen, Janne Hellsten, Timo Aila, Samuli Laine
Diffusion models currently dominate the field of data-driven image synthesis with their unparalleled scaling to large datasets. In this paper, we identify and rectify several causes for uneven and ineffective training in the popular ADM diffusion model architecture, without altering its high-level structure. Observing uncontrolled magnitude changes and imbalances in both the network activations and weights over the course of training, we redesign the network layers to preserve activation, weight, and update magnitudes on expectation. We find that systematic application of this philosophy eliminates the observed drifts and imbalances, resulting in considerably better networks at equal computational complexity. Our modifications improve the previous record FID of 2.41 in ImageNet-512 synthesis to 1.81, achieved using fast deterministic sampling.
As an independent contribution, we present a method for setting the exponential moving average (EMA) parameters post-hoc, i.e., after completing the training run. This allows precise tuning of EMA length without the cost of performing several training runs, and reveals its surprising interactions with network architecture, training time, and guidance.
Authors: Jan Schuchardt, Yan Scholten, Stephan Günnemann
A machine learning model is traditionally considered robust if its prediction remains (almost) constant under input perturbations with small norm. However, real-world tasks like molecular property prediction or point cloud segmentation have inherent equivariances, such as rotation or permutation equivariance. In such tasks, even perturbations with large norm do not necessarily change an input's semantic content. Furthermore, there are perturbations for which a model's prediction explicitly needs to change. For the first time, we propose a sound notion of adversarial robustness that accounts for task equivariance. We then demonstrate that provable robustness can be achieved by (1) choosing a model that matches the task's equivariances (2) certifying traditional adversarial robustness. Certification methods are, however, unavailable for many models, such as those with continuous equivariances. We close this gap by developing the framework of equivariance-preserving randomized smoothing, which enables architecture-agnostic certification. We additionally derive the first architecture-specific graph edit distance certificates, i.e. sound robustness guarantees for isomorphism equivariant tasks like node classification. Overall, a sound notion of robustness is an important prerequisite for future work at the intersection of robust and geometric machine learning.
Authors: Mingyu Huang, Ke Li
One of the most common problem-solving heuristics is by analogy. For a given problem, a solver can be viewed as a strategic walk on its fitness landscape. Thus if a solver works for one problem instance, we expect it will also be effective for other instances whose fitness landscapes essentially share structural similarities with each other. However, due to the black-box nature of combinatorial optimization, it is far from trivial to infer such similarity in real-world scenarios. To bridge this gap, by using local optima network as a proxy of fitness landscapes, this paper proposed to leverage graph data mining techniques to conduct qualitative and quantitative analyses to explore the latent topological structural information embedded in those landscapes. By conducting large-scale empirical experiments on three classic combinatorial optimization problems, we gain concrete evidence to support the existence of structural similarity between landscapes of the same classes within neighboring dimensions. We also interrogated the relationship between landscapes of different problem classes.
Authors: Max Klabunde, Mehdi Ben Amor, Michael Granitzer, Florian Lemmerich
Understanding the similarity of the numerous released large language models (LLMs) has many uses, e.g., simplifying model selection, detecting illegal model reuse, and advancing our understanding of what makes LLMs perform well. In this work, we measure the similarity of representations of a set of LLMs with 7B parameters. Our results suggest that some LLMs are substantially different from others. We identify challenges of using representational similarity measures that suggest the need of careful study of similarity scores to avoid false conclusions.
Authors: Kevin Badalian, Lucas Koch, Tobias Brinkmann, Mario Picerno, Marius Wegener, Sung-Yong Lee, Jakob Andert
Advances in artificial intelligence (AI) have led to its application in many areas of everyday life. In the context of control engineering, reinforcement learning (RL) represents a particularly promising approach as it is centred around the idea of allowing an agent to freely interact with its environment to find an optimal strategy. One of the challenges professionals face when training and deploying RL agents is that the latter often have to run on dedicated embedded devices. This could be to integrate them into an existing toolchain or to satisfy certain performance criteria like real-time constraints. Conventional RL libraries, however, cannot be easily utilised in conjunction with that kind of hardware. In this paper, we present a framework named LExCI, the Learning and Experiencing Cycle Interface, which bridges this gap and provides end-users with a free and open-source tool for training agents on embedded systems using the open-source library RLlib. Its operability is demonstrated with two state-of-the-art RL-algorithms and a rapid control prototyping system.
Authors: Xinnuo Xu, Ivan Titov, Mirella Lapata
Data-to-text generation involves transforming structured data, often represented as predicate-argument tuples, into coherent textual descriptions. Despite recent advances, systems still struggle when confronted with unseen combinations of predicates, producing unfaithful descriptions (e.g. hallucinations or omissions). We refer to this issue as compositional generalisation, and it encouraged us to create a benchmark for assessing the performance of different approaches on this specific problem. Furthermore, we propose a novel model that addresses compositional generalization by clustering predicates into groups. Our model generates text in a sentence-by-sentence manner, relying on one cluster of predicates at a time. This approach significantly outperforms T5~baselines across all evaluation metrics.Notably, it achieved a 31% improvement over T5 in terms of a metric focused on maintaining faithfulness to the input.
Authors: Hyunjik Kim, Matthias Bauer, Lucas Theis, Jonathan Richard Schwarz, Emilien Dupont
Most neural compression models are trained on large datasets of images or videos in order to generalize to unseen data. Such generalization typically requires large and expressive architectures with a high decoding complexity. Here we introduce C3, a neural compression method with strong rate-distortion (RD) performance that instead overfits a small model to each image or video separately. The resulting decoding complexity of C3 can be an order of magnitude lower than neural baselines with similar RD performance. C3 builds on COOL-CHIC (Ladune et al.) and makes several simple and effective improvements for images. We further develop new methodology to apply C3 to videos. On the CLIC2020 image benchmark, we match the RD performance of VTM, the reference implementation of the H.266 codec, with less than 3k MACs/pixel for decoding. On the UVG video benchmark, we match the RD performance of the Video Compression Transformer (Mentzer et al.), a well-established neural video codec, with less than 5k MACs/pixel for decoding.
Authors: Chenguang Zhao, Huan Yu
The macroscopic traffic flow model is widely used for traffic control and management. To incorporate drivers' anticipative behaviors and to remove impractical speed discontinuity inherent in the classic Lighthill-Whitham-Richards (LWR) traffic model, nonlocal partial differential equation (PDE) models with ``look-ahead" dynamics have been proposed, which assume that the speed is a function of weighted downstream traffic density. However, it lacks data validation on two important questions: whether there exist nonlocal dynamics, and how the length and weight of the ``look-ahead" window affect the spatial temporal propagation of traffic densities. In this paper, we adopt traffic trajectory data from a ring-road experiment and design a physics-informed neural network to learn the fundamental diagram and look-ahead kernel that best fit the data, and reinvent a data-enhanced nonlocal LWR model via minimizing the loss function combining the data discrepancy and the nonlocal model discrepancy. Results show that the learned nonlocal LWR yields a more accurate prediction of traffic wave propagation in three different scenarios: stop-and-go oscillations, congested, and free traffic. We first demonstrate the existence of ``look-ahead" effect with real traffic data. The optimal nonlocal kernel is found out to take a length of around 35 to 50 meters, and the kernel weight within 5 meters accounts for the majority of the nonlocal effect. Our results also underscore the importance of choosing a priori physics in machine learning models.
Authors: Stanislav Fort
We explore a class of adversarial attacks targeting the activations of language models. By manipulating a relatively small subset of model activations, $a$, we demonstrate the ability to control the exact prediction of a significant number (in some cases up to 1000) of subsequent tokens $t$. We empirically verify a scaling law where the maximum number of target tokens $t_\mathrm{max}$ predicted depends linearly on the number of tokens $a$ whose activations the attacker controls as $t_\mathrm{max} = \kappa a$. We find that the number of bits of control in the input space needed to control a single bit in the output space (what we call attack resistance $\chi$) is remarkably constant between $\approx 16$ and $\approx 25$ over 2 orders of magnitude of model sizes for different language models. Compared to attacks on tokens, attacks on activations are predictably much stronger, however, we identify a surprising regularity where one bit of input steered either via activations or via tokens is able to exert control over a similar amount of output bits. This gives support for the hypothesis that adversarial attacks are a consequence of dimensionality mismatch between the input and output spaces. A practical implication of the ease of attacking language model activations instead of tokens is for multi-modal and selected retrieval models, where additional data sources are added as activations directly, sidestepping the tokenized input. This opens up a new, broad attack surface. By using language models as a controllable test-bed to study adversarial attacks, we were able to experiment with input-output dimensions that are inaccessible in computer vision, especially where the output dimension dominates.
Authors: Bowen Jin, Gang Liu, Chi Han, Meng Jiang, Heng Ji, Jiawei Han
Large language models (LLMs), such as ChatGPT and LLaMA, are creating significant advancements in natural language processing, due to their strong text encoding/decoding ability and newly found emergent capability (e.g., reasoning). While LLMs are mainly designed to process pure texts, there are many real-world scenarios where text data are associated with rich structure information in the form of graphs (e.g., academic networks, and e-commerce networks) or scenarios where graph data are paired with rich textual information (e.g., molecules with descriptions). Besides, although LLMs have shown their pure text-based reasoning ability, it is underexplored whether such ability can be generalized to graph scenarios (i.e., graph-based reasoning). In this paper, we provide a systematic review of scenarios and techniques related to large language models on graphs. We first summarize potential scenarios of adopting LLMs on graphs into three categories, namely pure graphs, text-rich graphs, and text-paired graphs. We then discuss detailed techniques for utilizing LLMs on graphs, including LLM as Predictor, LLM as Encoder, and LLM as Aligner, and compare the advantages and disadvantages of different schools of models. Furthermore, we mention the real-world applications of such methods and summarize open-source codes and benchmark datasets. Finally, we conclude with potential future research directions in this fast-growing field. The related source can be found at https://github.com/PeterGriffinJin/Awesome-Language-Model-on-Graphs.
Authors: Sandeep Chinta, Xiang Gao, Qing Zhu
Methane (CH4) is the second most critical greenhouse gas after carbon dioxide, contributing to 16-25% of the observed atmospheric warming. Wetlands are the primary natural source of methane emissions globally. However, wetland methane emission estimates from biogeochemistry models contain considerable uncertainty. One of the main sources of this uncertainty arises from the numerous uncertain model parameters within various physical, biological, and chemical processes that influence methane production, oxidation, and transport. Sensitivity Analysis (SA) can help identify critical parameters for methane emission and achieve reduced biases and uncertainties in future projections. This study performs SA for 19 selected parameters responsible for critical biogeochemical processes in the methane module of the Energy Exascale Earth System Model (E3SM) land model (ELM). The impact of these parameters on various CH4 fluxes is examined at 14 FLUXNET- CH4 sites with diverse vegetation types. Given the extensive number of model simulations needed for global variance-based SA, we employ a machine learning (ML) algorithm to emulate the complex behavior of ELM methane biogeochemistry. ML enables the computational time to be shortened significantly from 6 CPU hours to 0.72 milliseconds, achieving reduced computational costs. We found that parameters linked to CH4 production and diffusion generally present the highest sensitivities despite apparent seasonal variation. Comparing simulated emissions from perturbed parameter sets against FLUXNET-CH4 observations revealed that better performances can be achieved at each site compared to the default parameter values. This presents a scope for further improving simulated emissions using parameter calibration with advanced optimization techniques like Bayesian optimization.
Authors: Yanjun Liu, Milena Jovanovic, Krishnanand Mallayya, Wesley J. Maddox, Andrew Gordon Wilson, Sebastian Klemenz, Leslie M. Schoop, Eun-Ah Kim
The advent of material databases provides an unprecedented opportunity to uncover predictive descriptors for emergent material properties from vast data space. However, common reliance on high-throughput ab initio data necessarily inherits limitations of such data: mismatch with experiments. On the other hand, experimental decisions are often guided by an expert's intuition honed from experiences that are rarely articulated. We propose using machine learning to "bottle" such operational intuition into quantifiable descriptors using expertly curated measurement-based data. We introduce "Materials Expert-Artificial Intelligence" (ME-AI) to encapsulate and articulate this human intuition. As a first step towards such a program, we focus on the topological semimetal (TSM) among square-net materials as the property inspired by the expert-identified descriptor based on structural information: the tolerance factor. We start by curating a dataset encompassing 12 primary features of 879 square-net materials, using experimental data whenever possible. We then use Dirichlet-based Gaussian process regression using a specialized kernel to reveal composite descriptors for square-net topological semimetals. The ME-AI learned descriptors independently reproduce expert intuition and expand upon it. Specifically, new descriptors point to hypervalency as a critical chemical feature predicting TSM within square-net compounds. Our success with a carefully defined problem points to the "machine bottling human insight" approach as promising for machine learning-aided material discovery.
Authors: Miriam Rateike, Celia Cintas, John Wamburu, Tanya Akumu, Skyler Speakman
We propose an auditing method to identify whether a large language model (LLM) encodes patterns such as hallucinations in its internal states, which may propagate to downstream tasks. We introduce a weakly supervised auditing technique using a subset scanning approach to detect anomalous patterns in LLM activations from pre-trained models. Importantly, our method does not need knowledge of the type of patterns a-priori. Instead, it relies on a reference dataset devoid of anomalies during testing. Further, our approach enables the identification of pivotal nodes responsible for encoding these patterns, which may offer crucial insights for fine-tuning specific sub-networks for bias mitigation. We introduce two new scanning methods to handle LLM activations for anomalous sentences that may deviate from the expected distribution in either direction. Our results confirm prior findings of BERT's limited internal capacity for encoding hallucinations, while OPT appears capable of encoding hallucination information internally. Importantly, our scanning approach, without prior exposure to false statements, performs comparably to a fully supervised out-of-distribution classifier.
Authors: Céline Comte, Matthieu Jonckheere, Jaron Sanders, Albert Senen-Cerda
Stochastic networks and queueing systems often lead to Markov decision processes (MDPs) with large state and action spaces as well as nonconvex objective functions, which hinders the convergence of many reinforcement learning (RL) algorithms. Policy-gradient methods perform well on MDPs with large state and action spaces, but they sometimes experience slow convergence due to the high variance of the gradient estimator. In this paper, we show that some of these difficulties can be circumvented by exploiting the structure of the underlying MDP. We first introduce a new family of gradient estimators called score-aware gradient estimators (SAGEs). When the stationary distribution of the MDP belongs to an exponential family parametrized by the policy parameters, SAGEs allow us to estimate the policy gradient without relying on value-function estimation, contrary to classical policy-gradient methods like actor-critic. To demonstrate their applicability, we examine two common control problems arising in stochastic networks and queueing systems whose stationary distributions have a product-form, a special case of exponential families. As a second contribution, we show that, under appropriate assumptions, the policy under a SAGE-based policy-gradient method has a large probability of converging to an optimal policy, provided that it starts sufficiently close to it, even with a nonconvex objective function and multiple maximizers. Our key assumptions are that, locally around a maximizer, a nondegeneracy property of the Hessian of the objective function holds and a Lyapunov function exists. Finally, we conduct a numerical comparison between a SAGE-based policy-gradient method and an actor-critic algorithm. The results demonstrate that the SAGE-based method finds close-to-optimal policies more rapidly, highlighting its superior performance over the traditional actor-critic method.
Authors: Florent Forest, Olga Fink
Intelligent Fault Diagnosis (IFD) based on deep learning has proven to be an effective and flexible solution, attracting extensive research. Deep neural networks can learn rich representations from vast amounts of representative labeled data for various applications. In IFD, they achieve high classification performance from signals in an end-to-end manner, without requiring extensive domain knowledge. However, deep learning models usually only perform well on the data distribution they have been trained on. When applied to a different distribution, they may experience performance drops. This is also observed in IFD, where assets are often operated in working conditions different from those in which labeled data have been collected. Unsupervised domain adaptation (UDA) deals with the scenario where labeled data are available in a source domain, and only unlabeled data are available in a target domain, where domains may correspond to operating conditions. Recent methods rely on training with confident pseudo-labels for target samples. However, the confidence-based selection of pseudo-labels is hindered by poorly calibrated confidence estimates in the target domain, primarily due to over-confident predictions, which limits the quality of pseudo-labels and leads to error accumulation. In this paper, we propose a novel UDA method called Calibrated Adaptive Teacher (CAT), where we propose to calibrate the predictions of the teacher network throughout the self-training process, leveraging post-hoc calibration techniques. We evaluate CAT on domain-adaptive IFD and perform extensive experiments on the Paderborn benchmark for bearing fault diagnosis under varying operating conditions. Our proposed method achieves state-of-the-art performance on most transfer tasks.
Authors: Rajeeva L. Karandikar, M. Vidyasagar
The Stochastic Approximation (SA) algorithm introduced by Robbins and Monro in 1951 has been a standard method for solving equations of the form $\mathbf{f}({\boldsymbol {\theta}}) = \mathbf{0}$, when only noisy measurements of $\mathbf{f}(\cdot)$ are available. If $\mathbf{f}({\boldsymbol {\theta}}) = \nabla J({\boldsymbol {\theta}})$ for some function $J(\cdot)$, then SA can also be used to find a stationary point of $J(\cdot)$. In much of the literature, it is assumed that the error term ${\boldsymbol {xi}}_{t+1}$ has zero conditional mean, and that its conditional variance is bounded as a function of $t$ (though not necessarily with respect to ${\boldsymbol {\theta}}_t$). Also, for the most part, the emphasis has been on ``synchronous'' SA, whereby, at each time $t$, \textit{every} component of ${\boldsymbol {\theta}}_t$ is updated. Over the years, SA has been applied to a variety of areas, out of which two are the focus in this paper: Convex and nonconvex optimization, and Reinforcement Learning (RL). As it turns out, in these applications, the above-mentioned assumptions do not always hold. In zero-order methods, the error neither has zero mean nor bounded conditional variance. In the present paper, we extend SA theory to encompass errors with nonzero conditional mean and/or unbounded conditional variance, and also asynchronous SA. In addition, we derive estimates for the rate of convergence of the algorithm. Then we apply the new results to problems in nonconvex optimization, and to Markovian SA, a recently emerging area in RL. We prove that SA converges in these situations, and compute the ``optimal step size sequences'' to maximize the estimated rate of convergence.
Authors: Nicolas Menet (1 and 2), Michael Hersche (1 and 2), Geethan Karunaratne (1), Luca Benini (2), Abu Sebastian (1), Abbas Rahimi (1) ((1) IBM Research - Zurich, (2) ETH Zurich)
With the advent of deep learning, progressively larger neural networks have been designed to solve complex tasks. We take advantage of these capacity-rich models to lower the cost of inference by exploiting computation in superposition. To reduce the computational burden per input, we propose Multiple-Input-Multiple-Output Neural Networks (MIMONets) capable of handling many inputs at once. MIMONets augment various deep neural network architectures with variable binding mechanisms to represent an arbitrary number of inputs in a compositional data structure via fixed-width distributed representations. Accordingly, MIMONets adapt nonlinear neural transformations to process the data structure holistically, leading to a speedup nearly proportional to the number of superposed input items in the data structure. After processing in superposition, an unbinding mechanism recovers each transformed input of interest. MIMONets also provide a dynamic trade-off between accuracy and throughput by an instantaneous on-demand switching between a set of accuracy-throughput operating points, yet within a single set of fixed parameters. We apply the concept of MIMONets to both CNN and Transformer architectures resulting in MIMOConv and MIMOFormer, respectively. Empirical evaluations show that MIMOConv achieves about 2-4 x speedup at an accuracy delta within [+0.68, -3.18]% compared to WideResNet CNNs on CIFAR10 and CIFAR100. Similarly, MIMOFormer can handle 2-4 inputs at once while maintaining a high average accuracy within a [-1.07, -3.43]% delta on the long range arena benchmark. Finally, we provide mathematical bounds on the interference between superposition channels in MIMOFormer. Our code is available at https://github.com/IBM/multiple-input-multiple-output-nets.
Authors: Lalit Pandey, Samantha M. W. Wood, Justin N. Wood
Vision transformers (ViTs) are top performing models on many computer vision benchmarks and can accurately predict human behavior on object recognition tasks. However, researchers question the value of using ViTs as models of biological learning because ViTs are thought to be more data hungry than brains, with ViTs requiring more training data to reach similar levels of performance. To test this assumption, we directly compared the learning abilities of ViTs and animals, by performing parallel controlled rearing experiments on ViTs and newborn chicks. We first raised chicks in impoverished visual environments containing a single object, then simulated the training data available in those environments by building virtual animal chambers in a video game engine. We recorded the first-person images acquired by agents moving through the virtual chambers and used those images to train self supervised ViTs that leverage time as a teaching signal, akin to biological visual systems. When ViTs were trained through the eyes of newborn chicks, the ViTs solved the same view invariant object recognition tasks as the chicks. Thus, ViTs were not more data hungry than newborn visual systems: both learned view invariant object representations in impoverished visual environments. The flexible and generic attention based learning mechanism in ViTs combined with the embodied data streams available to newborn animals appears sufficient to drive the development of animal-like object recognition.
Authors: Yiheng Jiang, Sinho Chewi, Aram-Alexandre Pooladian
We develop a theory of finite-dimensional polyhedral subsets over the Wasserstein space and optimization of functionals over them via first-order methods. Our main application is to the problem of mean-field variational inference, which seeks to approximate a distribution $\pi$ over $\mathbb{R}^d$ by a product measure $\pi^\star$. When $\pi$ is strongly log-concave and log-smooth, we provide (1) approximation rates certifying that $\pi^\star$ is close to the minimizer $\pi^\star_\diamond$ of the KL divergence over a \emph{polyhedral} set $\mathcal{P}_\diamond$, and (2) an algorithm for minimizing $\text{KL}(\cdot\|\pi)$ over $\mathcal{P}_\diamond$ with accelerated complexity $O(\sqrt \kappa \log(\kappa d/\varepsilon^2))$, where $\kappa$ is the condition number of $\pi$.
Authors: Tingting Hou, Chang Jiang, Qing Lu
The recent development of artificial intelligence (AI) technology, especially the advance of deep neural network (DNN) technology, has revolutionized many fields. While DNN plays a central role in modern AI technology, it has been rarely used in sequencing data analysis due to challenges brought by high-dimensional sequencing data (e.g., overfitting). Moreover, due to the complexity of neural networks and their unknown limiting distributions, building association tests on neural networks for genetic association analysis remains a great challenge. To address these challenges and fill the important gap of using AI in high-dimensional sequencing data analysis, we introduce a new kernel-based neural network (KNN) test for complex association analysis of sequencing data. The test is built on our previously developed KNN framework, which uses random effects to model the overall effects of high-dimensional genetic data and adopts kernel-based neural network structures to model complex genotype-phenotype relationships. Based on KNN, a Wald-type test is then introduced to evaluate the joint association of high-dimensional genetic data with a disease phenotype of interest, considering non-linear and non-additive effects (e.g., interaction effects). Through simulations, we demonstrated that our proposed method attained higher power compared to the sequence kernel association test (SKAT), especially in the presence of non-linear and interaction effects. Finally, we apply the methods to the whole genome sequencing (WGS) dataset from the Alzheimer's Disease Neuroimaging Initiative (ADNI) study, investigating new genes associated with the hippocampal volume change over time.
Authors: Tom Savage, Ehecatl Antonio del Rio Chanona
Domain experts often possess valuable physical insights that are overlooked in fully automated decision-making processes such as Bayesian optimisation. In this article we apply high-throughput (batch) Bayesian optimisation alongside anthropological decision theory to enable domain experts to influence the selection of optimal experiments. Our methodology exploits the hypothesis that humans are better at making discrete choices than continuous ones and enables experts to influence critical early decisions. At each iteration we solve an augmented multi-objective optimisation problem across a number of alternate solutions, maximising both the sum of their utility function values and the determinant of their covariance matrix, equivalent to their total variability. By taking the solution at the knee point of the Pareto front, we return a set of alternate solutions at each iteration that have both high utility values and are reasonably distinct, from which the expert selects one for evaluation. We demonstrate that even in the case of an uninformed practitioner, our algorithm recovers the regret of standard Bayesian optimisation.
Authors: Qiao Yu, Wengui Zhang, Jorge Cardoso, Odej Kao
In large-scale datacenters, memory failure is a common cause of server crashes, with uncorrectable errors (UEs) being a major indicator of Dual Inline Memory Module (DIMM) defects. Existing approaches primarily focus on predicting UEs using correctable errors (CEs), without fully considering the information provided by error bits. However, error bit patterns have a strong correlation with the occurrence of uncorrectable errors (UEs). In this paper, we present a comprehensive study on the correlation between CEs and UEs, specifically emphasizing the importance of spatio-temporal error bit information. Our analysis reveals a strong correlation between spatio-temporal error bits and UE occurrence. Through evaluations using real-world datasets, we demonstrate that our approach significantly improves prediction performance by 15% in F1-score compared to the state-of-the-art algorithms. Overall, our approach effectively reduces the number of virtual machine interruptions caused by UEs by approximately 59%.
Authors: Julien Boussard, Chandni Nagda, Julia Kaltenborn, Charlotte Emilie Elektra Lange, Philippe Brouillard, Yaniv Gurwicz, Peer Nowack, David Rolnick
Climate models, such as Earth system models (ESMs), are crucial for simulating future climate change based on projected Shared Socioeconomic Pathways (SSP) greenhouse gas emissions scenarios. While ESMs are sophisticated and invaluable, machine learning-based emulators trained on existing simulation data can project additional climate scenarios much faster and are computationally efficient. However, they often lack generalizability and interpretability. This work delves into the potential of causal representation learning, specifically the \emph{Causal Discovery with Single-parent Decoding} (CDSD) method, which could render climate model emulation efficient \textit{and} interpretable. We evaluate CDSD on multiple climate datasets, focusing on emissions, temperature, and precipitation. Our findings shed light on the challenges, limitations, and promise of using CDSD as a stepping stone towards more interpretable and robust climate model emulation.
Authors: Alexandra Zytek, Wei-En Wang, Sofia Koukoura, Kalyan Veeramachaneni
Through past experiences deploying what we call usable ML (one step beyond explainable ML, including both explanations and other augmenting information) to real-world domains, we have learned three key lessons. First, many organizations are beginning to hire people who we call ``bridges'' because they bridge the gap between ML developers and domain experts, and these people fill a valuable role in developing usable ML applications. Second, a configurable system that enables easily iterating on usable ML interfaces during collaborations with bridges is key. Finally, there is a need for continuous, in-deployment evaluations to quantify the real-world impact of usable ML. Throughout this paper, we apply these lessons to the task of wind turbine monitoring, an essential task in the renewable energy domain. Turbine engineers and data analysts must decide whether to perform costly in-person investigations on turbines to prevent potential cases of brakepad failure, and well-tuned usable ML interfaces can aid with this decision-making process. Through the applications of our lessons to this task, we hope to demonstrate the potential real-world impact of usable ML in the renewable energy domain.
Authors: Gaëtan Frusque, Ismail Nejjar, Majid Nabavi, Olga Fink
The Health Index (HI) is crucial for evaluating system health, aiding tasks like anomaly detection and predicting remaining useful life for systems demanding high safety and reliability. Tight monitoring is crucial for achieving high precision at a lower cost, with applications such as spray coating. Obtaining HI labels in real-world applications is often cost-prohibitive, requiring continuous, precise health measurements. Therefore, it is more convenient to leverage run-to failure datasets that may provide potential indications of machine wear condition, making it necessary to apply semi-supervised tools for HI construction. In this study, we adapt the Deep Semi-supervised Anomaly Detection (DeepSAD) method for HI construction. We use the DeepSAD embedding as a condition indicators to address interpretability challenges and sensitivity to system-specific factors. Then, we introduce a diversity loss to enrich condition indicators. We employ an alternating projection algorithm with isotonic constraints to transform the DeepSAD embedding into a normalized HI with an increasing trend. Validation on the PHME 2010 milling dataset, a recognized benchmark with ground truth HIs demonstrates meaningful HIs estimations. Our methodology is then applied to monitor wear states of thermal spray coatings using high-frequency voltage. Our contributions create opportunities for more accessible and reliable HI estimation, particularly in cases where obtaining ground truth HI labels is unfeasible.
Authors: Danyal Rehman, John H. Lienhard
Species transport models typically combine partial differential equations (PDEs) with relations from hindered transport theory to quantify electromigrative, convective, and diffusive transport through complex nanoporous systems; however, these formulations are frequently substantial simplifications of the governing dynamics, leading to the poor generalization performance of PDE-based models. Given the growing interest in deep learning methods for the physical sciences, we develop a machine learning-based approach to characterize ion transport across nanoporous membranes. Our proposed framework centers around attention-enhanced neural differential equations that incorporate electroneutrality-based inductive biases to improve generalization performance relative to conventional PDE-based methods. In addition, we study the role of the attention mechanism in illuminating physically-meaningful ion-pairing relationships across diverse mixture compositions. Further, we investigate the importance of pre-training on simulated data from PDE-based models, as well as the performance benefits from hard vs. soft inductive biases. Our results indicate that physics-informed deep learning solutions can outperform their classical PDE-based counterparts and provide promising avenues for modelling complex transport phenomena across diverse applications.
Authors: Angie Nataly Melo, Carlota Salinas, Miguel Angel Sotelo
In the context of autonomous driving, pedestrian crossing prediction is a key component for improving road safety. Presently, the focus of these predictions extends beyond achieving trustworthy results; it is shifting towards the explainability and interpretability of these predictions. This research introduces a novel neuro-symbolic approach that combines deep learning and fuzzy logic for an explainable and interpretable pedestrian crossing prediction. We have developed an explainable predictor (ExPedCross), which utilizes a set of explainable features and employs a fuzzy inference system to predict whether the pedestrian will cross or not. Our approach was evaluated on both the PIE and JAAD datasets. The results offer experimental insights into achieving explainability and interpretability in the pedestrian crossing prediction task. Furthermore, the testing results yield a set of guidelines and recommendations regarding the process of dataset selection, feature selection, and explainability.
Authors: Lukas Schulze Balhorn, Marc Caballero, Artur M. Schweidtmann
The process engineering domain widely uses Process Flow Diagrams (PFDs) and Process and Instrumentation Diagrams (P&IDs) to represent process flows and equipment configurations. However, the P&IDs and PFDs, hereafter called flowsheets, can contain errors causing safety hazards, inefficient operation, and unnecessary expenses. Correcting and verifying flowsheets is a tedious, manual process. We propose a novel generative AI methodology for automatically identifying errors in flowsheets and suggesting corrections to the user, i.e., autocorrecting flowsheets. Inspired by the breakthrough of Large Language Models (LLMs) for grammatical autocorrection of human language, we investigate LLMs for the autocorrection of flowsheets. The input to the model is a potentially erroneous flowsheet and the output of the model are suggestions for a corrected flowsheet. We train our autocorrection model on a synthetic dataset in a supervised manner. The model achieves a top-1 accuracy of 80% and a top-5 accuracy of 84% on an independent test dataset of synthetically generated flowsheets. The results suggest that the model can learn to autocorrect the synthetic flowsheets. We envision that flowsheet autocorrection will become a useful tool for chemical engineers.
Authors: Paolo Fazzini, Isaac Wheeler, Francesco Petracchini
In this work we analyze Multi-Agent Advantage Actor-Critic (MA2C) a recently proposed multi-agent reinforcement learning algorithm that can be applied to adaptive traffic signal control (ATSC) problems. To evaluate its potential we compare MA2C with Independent Advantage Actor-Critic (IA2C) and other Reinforcement Learning or heuristic based algorithms. Specifically, we analyze MA2C theoretically with the framework provided by non-Markov decision processes, which allows a deeper insight of the algorithm, and we critically examine the effectiveness and the robustness of the method by testing it in two traffic areas located in Bologna (Italy) simulated in SUMO, a software modeling tool for ATSC problems. Our results indicate that MA2C, trained with pseudo-random vehicle flows, is a promising technique able to outperform the alternative methods.
Authors: Haitao Mao, Lun Du, Yujia Zheng, Qiang Fu, Zelin Li, Xu Chen, Shi Han, Dongmei Zhang
Graph Neural Networks (GNNs) have achieved great success on a variety of tasks with graph-structural data, among which node classification is an essential one. Unsupervised Graph Domain Adaptation (UGDA) shows its practical value of reducing the labeling cost for node classification. It leverages knowledge from a labeled graph (i.e., source domain) to tackle the same task on another unlabeled graph (i.e., target domain). Most existing UGDA methods heavily rely on the labeled graph in the source domain. They utilize labels from the source domain as the supervision signal and are jointly trained on both the source graph and the target graph. However, in some real-world scenarios, the source graph is inaccessible because of privacy issues. Therefore, we propose a novel scenario named Source Free Unsupervised Graph Domain Adaptation (SFUGDA). In this scenario, the only information we can leverage from the source domain is the well-trained source model, without any exposure to the source graph and its labels. As a result, existing UGDA methods are not feasible anymore. To address the non-trivial adaptation challenges in this practical scenario, we propose a model-agnostic algorithm called SOGA for domain adaptation to fully exploit the discriminative ability of the source model while preserving the consistency of structural proximity on the target graph. We prove the effectiveness of the proposed algorithm both theoretically and empirically. The experimental results on four cross-domain tasks show consistent improvements in the Macro-F1 score and Macro-AUC.
Authors: Donghao Li, Yang Cao, Yuan Yao
Privacy-preserving data release algorithms have gained increasing attention for their ability to protect user privacy while enabling downstream machine learning tasks. However, the utility of current popular algorithms is not always satisfactory. Mixup of raw data provides a new way of data augmentation, which can help improve utility. However, its performance drastically deteriorates when differential privacy (DP) noise is added. To address this issue, this paper draws inspiration from the recently observed Neural Collapse (NC) phenomenon, which states that the last layer features of a neural network concentrate on the vertices of a simplex as Equiangular Tight Frame (ETF). We propose a scheme to mixup the Neural Collapse features to exploit the ETF simplex structure and release noisy mixed features to enhance the utility of the released data. By using Gaussian Differential Privacy (GDP), we obtain an asymptotic rate for the optimal mixup degree. To further enhance the utility and address the label collapse issue when the mixup degree is large, we propose a Hierarchical sampling method to stratify the mixup samples on a small number of classes. This method remarkably improves utility when the number of classes is large. Extensive experiments demonstrate the effectiveness of our proposed method in protecting against attacks and improving utility. In particular, our approach shows significantly improved utility compared to directly training classification networks with DPSGD on CIFAR100 and MiniImagenet datasets, highlighting the benefits of using privacy-preserving data release. We release reproducible code in https://github.com/Lidonghao1996/NeuroMixGDP.
Authors: Yuan Zhang, Jianhong Wang, Joschka Boedecker
Reinforcement learning (RL) is recognized as lacking generalization and robustness under environmental perturbations, which excessively restricts its application for real-world robotics. Prior work claimed that adding regularization to the value function is equivalent to learning a robust policy with uncertain transitions. Although the regularization-robustness transformation is appealing for its simplicity and efficiency, it is still lacking in continuous control tasks. In this paper, we propose a new regularizer named $\textbf{U}$ncertainty $\textbf{S}$et $\textbf{R}$egularizer (USR), by formulating the uncertainty set on the parameter space of the transition function. In particular, USR is flexible enough to be plugged into any existing RL framework. To deal with unknown uncertainty sets, we further propose a novel adversarial approach to generate them based on the value function. We evaluate USR on the Real-world Reinforcement Learning (RWRL) benchmark, demonstrating improvements in the robust performance for perturbed testing environments.
Authors: Qin Yang, Ramviyas Parasuraman
Adopting reasonable strategies is challenging but crucial for an intelligent agent with limited resources working in hazardous, unstructured, and dynamic environments to improve the system's utility, decrease the overall cost, and increase mission success probability. This paper proposes a novel directed acyclic strategy graph decomposition approach based on Bayesian chaining to separate an intricate policy into several simple sub-policies and organize their relationships as Bayesian strategy networks (BSN). We integrate this approach into the state-of-the-art DRL method -- soft actor-critic (SAC), and build the corresponding Bayesian soft actor-critic (BSAC) model by organizing several sub-policies as a joint policy. We compare our method against the state-of-the-art deep reinforcement learning algorithms on the standard continuous control benchmarks in the OpenAI Gym environment. The results demonstrate that the promising potential of the BSAC method significantly improves training efficiency.
Authors: Ryoma Yataka, Kazuki Hirashima, Masashi Shiraishi
Recently, studies on machine learning have focused on methods that use symmetry implicit in a specific manifold as an inductive bias. Grassmann manifolds provide the ability to handle fundamental shapes represented as shape spaces, enabling stable shape analysis. In this paper, we present a novel approach in which we establish the theoretical foundations for learning distributions on the Grassmann manifold via continuous normalization flows, with the explicit goal of generating stable shapes. Our approach facilitates more robust generation by effectively eliminating the influence of extraneous transformations, such as rotations and inversions, through learning and generating within a Grassmann manifold designed to accommodate the essential shape information of the object. The experimental results indicated that the proposed method could generate high-quality samples by capturing the data structure. Furthermore, the proposed method significantly outperformed state-of-the-art methods in terms of the log-likelihood or evidence lower bound. The results obtained are expected to stimulate further research in this field, leading to advances for stable shape generation and analysis.
Authors: Philipp Pilar, Niklas Wahlström
Physics-informed neural networks (PINNs) constitute a flexible approach to both finding solutions and identifying parameters of partial differential equations. Most works on the topic assume noiseless data, or data contaminated with weak Gaussian noise. We show that the standard PINN framework breaks down in case of non-Gaussian noise. We give a way of resolving this fundamental issue and we propose to jointly train an energy-based model (EBM) to learn the correct noise distribution. We illustrate the improved performance of our approach using multiple examples.
Authors: Yihan Wu, Junliang Guo, Xu Tan, Chen Zhang, Bohan Li, Ruihua Song, Lei He, Sheng Zhao, Arul Menezes, Jiang Bian
Video dubbing aims to translate the original speech in a film or television program into the speech in a target language, which can be achieved with a cascaded system consisting of speech recognition, machine translation and speech synthesis. To ensure the translated speech to be well aligned with the corresponding video, the length/duration of the translated speech should be as close as possible to that of the original speech, which requires strict length control. Previous works usually control the number of words or characters generated by the machine translation model to be similar to the source sentence, without considering the isochronicity of speech as the speech duration of words/characters in different languages varies. In this paper, we propose a machine translation system tailored for the task of video dubbing, which directly considers the speech duration of each token in translation, to match the length of source and target speech. Specifically, we control the speech length of generated sentence by guiding the prediction of each word with the duration information, including the speech duration of itself as well as how much duration is left for the remaining words. We design experiments on four language directions (German -> English, Spanish -> English, Chinese <-> English), and the results show that the proposed method achieves better length control ability on the generated speech than baseline methods. To make up the lack of real-world datasets, we also construct a real-world test set collected from films to provide comprehensive evaluations on the video dubbing task.
Authors: Martin Hellkvist, Ayça Özçelikkale, Anders Ahlén
We focus on the continual learning problem where the tasks arrive sequentially and the aim is to perform well on the newly arrived task without performance degradation on the previously seen tasks. In contrast to the continual learning literature focusing on the centralized setting, we investigate the distributed estimation framework. We consider the well-established distributed learning algorithm COCOA. We derive closed form expressions for the iterations for the overparametrized case. We illustrate the convergence and the error performance of the algorithm based on the over/under-parameterization of the problem. Our results show that depending on the problem dimensions and data generation assumptions, COCOA can perform continual learning over a sequence of tasks, i.e., it can learn a new task without forgetting previously learned tasks, with access only to one task at a time.
Authors: Martin Hellkvist, Ayça Özçelikkale, Anders Ahlén
Recent successes of massively overparameterized models have inspired a new line of work investigating the underlying conditions that enable overparameterized models to generalize well. This paper considers a framework where the possibly overparametrized model includes fake features, i.e., features that are present in the model but not in the data. We present a non-asymptotic high-probability bound on the generalization error of the ridge regression problem under the model misspecification of having fake features. Our highprobability results provide insights into the interplay between the implicit regularization provided by the fake features and the explicit regularization provided by the ridge parameter. Numerical results illustrate the trade-off between the number of fake features and how the optimal ridge parameter may heavily depend on the number of fake features.
Authors: Ethan Steinberg, Jason Fries, Yizhe Xu, Nigam Shah
We present a self-supervised, time-to-event (TTE) foundation model called MOTOR (Many Outcome Time Oriented Representations) which is pretrained on timestamped sequences of events in electronic health records (EHR) and health insurance claims. TTE models are used for estimating the probability distribution of the time until a specific event occurs, which is an important task in medical settings. TTE models provide many advantages over classification using fixed time horizons, including naturally handling censored observations, but are challenging to train with limited labeled data. MOTOR addresses this challenge by pretraining on up to 55M patient records (9B clinical events). We evaluate MOTOR's transfer learning performance on 19 tasks, across 3 patient databases (a private EHR system, MIMIC-IV, and Merative claims data). Task-specific models adapted from MOTOR improve time-dependent C statistics by 4.6% over state-of-the-art, improve label efficiency by up to 95% ,and are more robust to temporal distributional shifts. We further evaluate cross-site portability by adapting our MOTOR foundation model for six prediction tasks on the MIMIC-IV dataset, where it outperforms all baselines. MOTOR is the first foundation model for medical TTE predictions and we release a 143M parameter pretrained model for research use at [redacted URL].
Authors: Bowen Lei, Dongkuan Xu, Ruqi Zhang, Shuren He, Bani K. Mallick
Despite impressive performance, deep neural networks require significant memory and computation costs, prohibiting their application in resource-constrained scenarios. Sparse training is one of the most common techniques to reduce these costs, however, the sparsity constraints add difficulty to the optimization, resulting in an increase in training time and instability. In this work, we aim to overcome this problem and achieve space-time co-efficiency. To accelerate and stabilize the convergence of sparse training, we analyze the gradient changes and develop an adaptive gradient correction method. Specifically, we approximate the correlation between the current and previous gradients, which is used to balance the two gradients to obtain a corrected gradient. Our method can be used with the most popular sparse training pipelines under both standard and adversarial setups. Theoretically, we prove that our method can accelerate the convergence rate of sparse training. Extensive experiments on multiple datasets, model architectures, and sparsities demonstrate that our method outperforms leading sparse training methods by up to \textbf{5.0\%} in accuracy given the same number of training epochs, and reduces the number of training epochs by up to \textbf{52.1\%} to achieve the same accuracy. Our code is available on: \url{https://github.com/StevenBoys/AGENT}.
Authors: Danny Wood, Tingting Mu, Andrew Webb, Henry Reeve, Mikel Lujan, Gavin Brown
We present a theory of ensemble diversity, explaining the nature of diversity for a wide range of supervised learning scenarios. This challenge, of understanding ensemble diversity, has been referred to as the "holy grail" of ensemble learning, an open research issue for over 30 years. Our framework reveals that diversity is in fact a hidden dimension in the bias-variance decomposition of the ensemble loss. We prove a family of exact bias-variance-diversity decompositions, for both regression and classification, e.g., squared, cross-entropy, and Poisson losses. For losses where an additive bias-variance decomposition is not available (e.g., 0/1 loss) we present an alternative approach, which precisely quantifies the effects of diversity, turning out to be dependent on the label distribution. Experiments show how we can use our framework to understand the diversity-encouraging mechanisms of popular methods: Bagging, Boosting, and Random Forests.
Authors: Benjamin J. Holzschuh, Simona Vegetti, Nils Thuerey
We propose to solve inverse problems involving the temporal evolution of physics systems by leveraging recent advances from diffusion models. Our method moves the system's current state backward in time step by step by combining an approximate inverse physics simulator and a learned correction function. A central insight of our work is that training the learned correction with a single-step loss is equivalent to a score matching objective, while recursively predicting longer parts of the trajectory during training relates to maximum likelihood training of a corresponding probability flow. We highlight the advantages of our algorithm compared to standard denoising score matching and implicit score matching, as well as fully learned baselines for a wide range of inverse physics problems. The resulting inverse solver has excellent accuracy and temporal stability and, in contrast to other learned inverse solvers, allows for sampling the posterior of the solutions.
Authors: Dovile Juodelyte, Amelia Jiménez-Sánchez, Veronika Cheplygina
While a key component to the success of deep learning is the availability of massive amounts of training data, medical image datasets are often limited in diversity and size. Transfer learning has the potential to bridge the gap between related yet different domains. For medical applications, however, it remains unclear whether it is more beneficial to pre-train on natural or medical images. We aim to shed light on this problem by comparing initialization on ImageNet and RadImageNet on seven medical classification tasks. Our work includes a replication study, which yields results contrary to previously published findings. In our experiments, ResNet50 models pre-trained on ImageNet tend to outperform those trained on RadImageNet. To gain further insights, we investigate the learned representations using Canonical Correlation Analysis (CCA) and compare the predictions of the different models. Our results indicate that, contrary to intuition, ImageNet and RadImageNet may converge to distinct intermediate representations, which appear to diverge further during fine-tuning. Despite these distinct representations, the predictions of the models remain similar. Our findings show that the similarity between networks before and after fine-tuning does not correlate with performance gains, suggesting that the advantages of transfer learning might not solely originate from the reuse of features in the early layers of a convolutional neural network.
Authors: Alessio Capitanelli, Fulvio Mastrogiovanni
Symbolic task planning is a widely used approach to enforce robot autonomy due to its ease of understanding and deployment. However, symbolic task planning is difficult to scale in real-world when frequent re-planning is needed, for example, due to human-robot interactions or unforeseen events. Plan length and planning time can hinder the robot's efficiency and negatively affect the overall human-robot interaction's fluency. We present a framework, Teriyaki, designed to bridge the gap between symbolic task planning and machine learning approaches, by training Large Language Models (LLMs), namely GPT-3, into neurosymbolic task planners compatible with the Planning Domain Definition Language (PDDL). Potential benefits include: (i) better scalability in so far as the planning domain complexity increases, since LLMs' response time linearly scales with the combined length of the input and the output, instead of super-linearly as in the case of symbolic task planners, and (ii) the ability to synthesize a plan action-by-action instead of end-to-end, and to make each action available for execution as soon as it is generated, which in turn enables concurrent planning and execution. In the past year, significant efforts have been devoted by the research community to evaluate the overall cognitive abilities of LLMs, with alternate successes. Instead, with Teriyaki we aim to providing an overall planning performance comparable to traditional planners in specific planning domains, while leveraging LLMs capabilities in other metrics which are used to build a look-ahead predictive planning model. Preliminary results in selected domains show that our method can: (i) solve 95.5% of problems in a test data set of 1000 samples; (ii) produce plans up to 13.5% shorter than a traditional symbolic planner; (iii) reduce average overall waiting times for a plan availability by up to 61.4%.
Authors: Zhenmei Shi, Yifei Ming, Ying Fan, Frederic Sala, Yingyu Liang
The ability to generalize to unseen domains is crucial for machine learning systems deployed in the real world, especially when we only have data from limited training domains. In this paper, we propose a simple and effective regularization method based on the nuclear norm of the learned features for domain generalization. Intuitively, the proposed regularizer mitigates the impacts of environmental features and encourages learning domain-invariant features. Theoretically, we provide insights into why nuclear norm regularization is more effective compared to ERM and alternative regularization methods. Empirically, we conduct extensive experiments on both synthetic and real datasets. We show nuclear norm regularization achieves strong performance compared to baselines in a wide range of domain generalization tasks. Moreover, our regularizer is broadly applicable with various methods such as ERM and SWAD with consistently improved performance, e.g., 1.7% and 0.9% test accuracy improvements respectively on the DomainBed benchmark.
Authors: Chen Xie, Francesco Daghero, Yukai Chen, Marco Castellano, Luca Gandolfi, Andrea Calimera, Enrico Macii, Massimo Poncino, Daniele Jahier Pagliari
Ultra-low-resolution Infrared (IR) array sensors offer a low-cost, energy-efficient, and privacy-preserving solution for people counting, with applications such as occupancy monitoring. Previous work has shown that Deep Learning (DL) can yield superior performance on this task. However, the literature was missing an extensive comparative analysis of various efficient DL architectures for IR array-based people counting, that considers not only their accuracy, but also the cost of deploying them on memory- and energy-constrained Internet of Things (IoT) edge nodes. In this work, we address this need by comparing 6 different DL architectures on a novel dataset composed of IR images collected from a commercial 8x8 array, which we made openly available. With a wide architectural exploration of each model type, we obtain a rich set of Pareto-optimal solutions, spanning cross-validated balanced accuracy scores in the 55.70-82.70% range. When deployed on a commercial Microcontroller (MCU) by STMicroelectronics, the STM32L4A6ZG, these models occupy 0.41-9.28kB of memory, and require 1.10-7.74ms per inference, while consuming 17.18-120.43 $\mu$J of energy. Our models are significantly more accurate than a previous deterministic method (up to +39.9%), while being up to 3.53x faster and more energy efficient. Further, our models' accuracy is comparable to state-of-the-art DL solutions on similar resolution sensors, despite a much lower complexity. All our models enable continuous, real-time inference on a MCU-based IoT node, with years of autonomous operation without battery recharging.
Authors: Philipp Altmann, Fabian Ritz, Leonard Feuchtinger, Jonas Nüßlein, Claudia Linnhoff-Popien, Thomy Phan
The safe application of reinforcement learning (RL) requires generalization from limited training data to unseen scenarios. Yet, fulfilling tasks under changing circumstances is a key challenge in RL. Current state-of-the-art approaches for generalization apply data augmentation techniques to increase the diversity of training data. Even though this prevents overfitting to the training environment(s), it hinders policy optimization. Crafting a suitable observation, only containing crucial information, has been shown to be a challenging task itself. To improve data efficiency and generalization capabilities, we propose Compact Reshaped Observation Processing (CROP) to reduce the state information used for policy optimization. By providing only relevant information, overfitting to a specific training layout is precluded and generalization to unseen environments is improved. We formulate three CROPs that can be applied to fully observable observation- and action-spaces and provide methodical foundation. We empirically show the improvements of CROP in a distributionally shifted safety gridworld. We furthermore provide benchmark comparisons to full observability and data-augmentation in two different-sized procedurally generated mazes.
Authors: Kevin Zeng, Carlos E. Pérez De Jesús, Andrew J. Fox, Michael D. Graham
While many phenomena in physics and engineering are formally high-dimensional, their long-time dynamics often live on a lower-dimensional manifold. The present work introduces an autoencoder framework that combines implicit regularization with internal linear layers and $L_2$ regularization (weight decay) to automatically estimate the underlying dimensionality of a data set, produce an orthogonal manifold coordinate system, and provide the mapping functions between the ambient space and manifold space, allowing for out-of-sample projections. We validate our framework's ability to estimate the manifold dimension for a series of datasets from dynamical systems of varying complexities and compare to other state-of-the-art estimators. We analyze the training dynamics of the network to glean insight into the mechanism of low-rank learning and find that collectively each of the implicit regularizing layers compound the low-rank representation and even self-correct during training. Analysis of gradient descent dynamics for this architecture in the linear case reveals the role of the internal linear layers in leading to faster decay of a "collective weight variable" incorporating all layers, and the role of weight decay in breaking degeneracies and thus driving convergence along directions in which no decay would occur in its absence. We show that this framework can be naturally extended for applications of state-space modeling and forecasting by generating a data-driven dynamic model of a spatiotemporally chaotic partial differential equation using only the manifold coordinates. Finally, we demonstrate that our framework is robust to hyperparameter choices.
Authors: Alexandre Maraval, Matthieu Zimmer, Antoine Grosnit, Haitham Bou Ammar
Meta-Bayesian optimisation (meta-BO) aims to improve the sample efficiency of Bayesian optimisation by leveraging data from related tasks. While previous methods successfully meta-learn either a surrogate model or an acquisition function independently, joint training of both components remains an open challenge. This paper proposes the first end-to-end differentiable meta-BO framework that generalises neural processes to learn acquisition functions via transformer architectures. We enable this end-to-end framework with reinforcement learning (RL) to tackle the lack of labelled acquisition data. Early on, we notice that training transformer-based neural processes from scratch with RL is challenging due to insufficient supervision, especially when rewards are sparse. We formalise this claim with a combinatorial analysis showing that the widely used notion of regret as a reward signal exhibits a logarithmic sparsity pattern in trajectory lengths. To tackle this problem, we augment the RL objective with an auxiliary task that guides part of the architecture to learn a valid probabilistic model as an inductive bias. We demonstrate that our method achieves state-of-the-art regret results against various baselines in experiments on standard hyperparameter optimisation tasks and also outperforms others in the real-world problems of mixed-integer programming tuning, antibody design, and logic synthesis for electronic design automation.
Authors: Naoya Hasegawa, Issei Sato
Recognition problems in long-tailed data, in which the sample size per class is heavily skewed, have gained importance because the distribution of the sample size per class in a dataset is generally exponential unless the sample size is intentionally adjusted. Various methods have been devised to address these problems. Recently, weight balancing, which combines well-known classical regularization techniques with two-stage training, has been proposed. Despite its simplicity, it is known for its high performance compared with existing methods devised in various ways. However, there is a lack of understanding as to why this method is effective for long-tailed data. In this study, we analyze weight balancing by focusing on neural collapse and the cone effect at each training stage and found that it can be decomposed into an increase in Fisher's discriminant ratio of the feature extractor caused by weight decay and cross entropy loss and implicit logit adjustment caused by weight decay and class-balanced loss. Our analysis enables the training method to be further simplified by reducing the number of training stages to one while increasing accuracy.
Authors: Yingyi Chen, Qinghua Tao, Francesco Tonin, Johan A.K. Suykens
Recently, a new line of works has emerged to understand and improve self-attention in Transformers by treating it as a kernel machine. However, existing works apply the methods for symmetric kernels to the asymmetric self-attention, resulting in a nontrivial gap between the analytical understanding and numerical implementation. In this paper, we provide a new perspective to represent and optimize self-attention through asymmetric Kernel Singular Value Decomposition (KSVD), which is also motivated by the low-rank property of self-attention normally observed in deep layers. Through asymmetric KSVD, $i$) a primal-dual representation of self-attention is formulated, where the optimization objective is cast to maximize the projection variances in the attention outputs; $ii$) a novel attention mechanism, i.e., Primal-Attention, is proposed via the primal representation of KSVD, avoiding explicit computation of the kernel matrix in the dual; $iii$) with KKT conditions, we prove that the stationary solution to the KSVD optimization in Primal-Attention yields a zero-value objective. In this manner, KSVD optimization can be implemented by simply minimizing a regularization loss, so that low-rank property is promoted without extra decomposition. Numerical experiments show state-of-the-art performance of our Primal-Attention with improved efficiency. Moreover, we demonstrate that the deployed KSVD optimization regularizes Primal-Attention with a sharper singular value decay than that of the canonical self-attention, further verifying the great potential of our method. To the best of our knowledge, this is the first work that provides a primal-dual representation for the asymmetric kernel in self-attention and successfully applies it to modeling and optimization.
Authors: Daniel C.H. Tan, Fernando Acero, Robert McCarthy, Dimitrios Kanoulas, Zhibin Li
Guaranteeing safe behaviour of reinforcement learning (RL) policies poses significant challenges for safety-critical applications, despite RL's generality and scalability. To address this, we propose a new approach to apply verification methods from control theory to learned value functions. By analyzing task structures for safety preservation, we formalize original theorems that establish links between value functions and control barrier functions. Further, we propose novel metrics for verifying value functions in safe control tasks and practical implementation details to improve learning. Our work presents a novel method for certificate learning, which unlocks a diversity of verification techniques from control theory for RL policies, and marks a significant step towards a formal framework for the general, scalable, and verifiable design of RL-based control systems. Code and videos are available at this https url: https://rl-cbf.github.io/
Authors: Akshatha Mohan, Joshua Peeples
We present a comprehensive analysis of quantitatively evaluating explainable artificial intelligence (XAI) techniques for remote sensing image classification. Our approach leverages state-of-the-art machine learning approaches to perform remote sensing image classification across multiple modalities. We investigate the results of the models qualitatively through XAI methods. Additionally, we compare the XAI methods quantitatively through various categories of desired properties. Through our analysis, we offer insights and recommendations for selecting the most appropriate XAI method(s) to gain a deeper understanding of the models' decision-making processes. The code for this work is publicly available.
Authors: Enrique Tomás Martínez Beltrán, Ángel Luis Perales Gómez, Chao Feng, Pedro Miguel Sánchez Sánchez, Sergio López Bernal, Gérôme Bovet, Manuel Gil Pérez, Gregorio Martínez Pérez, Alberto Huertas Celdrán
In 2016, Google proposed Federated Learning (FL) as a novel paradigm to train Machine Learning (ML) models across the participants of a federation while preserving data privacy. Since its birth, Centralized FL (CFL) has been the most used approach, where a central entity aggregates participants' models to create a global one. However, CFL presents limitations such as communication bottlenecks, single point of failure, and reliance on a central server. Decentralized Federated Learning (DFL) addresses these issues by enabling decentralized model aggregation and minimizing dependency on a central entity. Despite these advances, current platforms training DFL models struggle with key issues such as managing heterogeneous federation network topologies. To overcome these challenges, this paper presents Fedstellar, a novel platform designed to train FL models in a decentralized, semi-decentralized, and centralized fashion across diverse federations of physical or virtualized devices. The Fedstellar implementation encompasses a web application with an interactive graphical interface, a controller for deploying federations of nodes using physical or virtual devices, and a core deployed on each device which provides the logic needed to train, aggregate, and communicate in the network. The effectiveness of the platform has been demonstrated in two scenarios: a physical deployment involving single-board devices such as Raspberry Pis for detecting cyberattacks, and a virtualized deployment comparing various FL approaches in a controlled environment using MNIST and CIFAR-10 datasets. In both scenarios, Fedstellar demonstrated consistent performance and adaptability, achieving F1 scores of 91%, 98%, and 91.2% using DFL for detecting cyberattacks and classifying MNIST and CIFAR-10, respectively, reducing training time by 32% compared to centralized approaches.
Authors: Matthew Wallingford, Vivek Ramanujan, Alex Fang, Aditya Kusupati, Roozbeh Mottaghi, Aniruddha Kembhavi, Ludwig Schmidt, Ali Farhadi
We propose Neural Priming, a technique for adapting large pretrained models to distribution shifts and downstream tasks given few or no labeled examples. Presented with class names or unlabeled test samples, Neural Priming enables the model to recall and conditions its parameters on relevant data seen throughout pretraining, thereby priming it for the test distribution. Neural Priming can be performed at test time, even for pretraining datasets as large as LAION-2B. Performing lightweight updates on the recalled data significantly improves accuracy across a variety of distribution shift and transfer learning benchmarks. Concretely, in the zero-shot setting, we see a 2.45% improvement in accuracy on ImageNet and 3.81% accuracy improvement on average across standard transfer learning benchmarks. Further, using Neural Priming at inference to adapt to distribution shift, we see a 1.41% accuracy improvement on ImageNetV2. These results demonstrate the effectiveness of Neural Priming in addressing the challenge of limited labeled data and changing distributions. Code is available at github.com/RAIVNLab/neural-priming.
Authors: Ajit Mahata, Reetish Padhi, Amit Apte
Study of dynamical systems using partial state observation is an important problem due to its applicability to many real-world systems. We address the problem by studying an echo state network (ESN) framework with partial state input with partial or full state output. Application to the Lorenz system and Chua's oscillator (both numerically simulated and experimental systems) demonstrate the effectiveness of our method. We show that the ESN, as an autonomous dynamical system, is capable of making short-term predictions up to a few Lyapunov times. However, the prediction horizon has high variability depending on the initial condition-an aspect that we explore in detail using the distribution of the prediction horizon. Further, using a variety of statistical metrics to compare the long-term dynamics of the ESN predictions with numerically simulated or experimental dynamics and observed similar results, we show that the ESN can effectively learn the system's dynamics even when trained with noisy numerical or experimental datasets. Thus, we demonstrate the potential of ESNs to serve as cheap surrogate models for simulating the dynamics of systems where complete observations are unavailable.
Authors: Vitória Barin-Pacela, Kartik Ahuja, Simon Lacoste-Julien, Pascal Vincent
Disentanglement aims to recover meaningful latent ground-truth factors from the observed distribution solely, and is formalized through the theory of identifiability. The identifiability of independent latent factors is proven to be impossible in the unsupervised i.i.d. setting under a general nonlinear map from factors to observations. In this work, however, we demonstrate that it is possible to recover quantized latent factors under a generic nonlinear diffeomorphism. We only assume that the latent factors have independent discontinuities in their density, without requiring the factors to be statistically independent. We introduce this novel form of identifiability, termed quantized factor identifiability, and provide a comprehensive proof of the recovery of the quantized factors.
Authors: Yaswitha Gujju, Atsushi Matsuo, Rudy Raymond
The past decade has witnessed significant advancements in quantum hardware, encompassing improvements in speed, qubit quantity, and quantum volume-a metric defining the maximum size of a quantum circuit effectively implementable on near-term quantum devices. This progress has led to a surge in Quantum Machine Learning (QML) applications on real hardware, aiming to achieve quantum advantage over classical approaches. This survey focuses on selected supervised and unsupervised learning applications executed on quantum hardware, specifically tailored for real-world scenarios. The exploration includes a thorough analysis of current QML implementation limitations on quantum hardware, covering techniques like encoding, ansatz structure, error mitigation, and gradient methods to address these challenges. Furthermore, the survey evaluates the performance of QML implementations in comparison to classical counterparts. In conclusion, we discuss existing bottlenecks related to applying QML on real quantum devices and propose potential solutions to overcome these challenges in the future.
Authors: Jincheng Zhou, Beatrice Bevilacqua, Bruno Ribeiro
The task of inductive link prediction in (discrete) attributed multigraphs infers missing attributed links (relations) between nodes in new test multigraphs. Traditional relational learning methods face the challenge of limited generalization to test multigraphs containing both novel nodes and novel relation types not seen in training. Recently, under the only assumption that all relation types share the same structural predictive patterns (single task), Gao et al. (2023) proposed a link prediction method using the theoretical concept of double equivariance (equivariance for nodes & relation types), in contrast to the (single) equivariance (only for nodes) used to design Graph Neural Networks (GNNs). In this work we further extend the double equivariance concept to multi-task double equivariance, where we define link prediction in attributed multigraphs that can have distinct and potentially conflicting predictive patterns for different sets of relation types (multiple tasks). Our empirical results on real-world datasets demonstrate that our approach can effectively generalize to test graphs with multi-task structures without access to additional information.
Authors: Xingrong Dong, Zhaoxian Wu, Qing Ling, Zhi Tian
This paper studies distributed online learning under Byzantine attacks. The performance of an online learning algorithm is often characterized by (adversarial) regret, which evaluates the quality of one-step-ahead decision-making when an environment provides adversarial losses, and a sublinear bound is preferred. But we prove that, even with a class of state-of-the-art robust aggregation rules, in an adversarial environment and in the presence of Byzantine participants, distributed online gradient descent can only achieve a linear adversarial regret bound, which is tight. This is the inevitable consequence of Byzantine attacks, even though we can control the constant of the linear adversarial regret to a reasonable level. Interestingly, when the environment is not fully adversarial so that the losses of the honest participants are i.i.d. (independent and identically distributed), we show that sublinear stochastic regret, in contrast to the aforementioned adversarial regret, is possible. We develop a Byzantine-robust distributed online momentum algorithm to attain such a sublinear stochastic regret bound. Extensive numerical experiments corroborate our theoretical analysis.
Authors: Paolo Climaco, Jochen Garcke
For regression tasks one often leverages large datasets for training predictive machine learning models. However, using large datasets may not be feasible due to computational limitations or high data labelling costs. Therefore, suitably selecting small training sets from large pools of unlabelled data points is essential to maximize model performance while maintaining efficiency. In this work, we study Farthest Point Sampling (FPS), a data selection approach that aims to minimize the fill distance of the selected set. We derive an upper bound for the maximum expected prediction error, conditional to the location of the unlabelled data points, that linearly depends on the training set fill distance. For empirical validation, we perform experiments using two regression models on three datasets. We empirically show that selecting a training set by aiming to minimize the fill distance, thereby minimizing our derived bound, significantly reduces the maximum prediction error of various regression models, outperforming alternative sampling approaches by a large margin. Furthermore, we show that selecting training sets with the FPS can also increase model stability for the specific case of Gaussian kernel regression approaches.
Authors: Xinyue Ma, Suyeon Jeong, Minjia Zhang, Di Wang, Jonghyun Choi, Myeongjae Jeon
Continual learning (CL) trains NN models incrementally from a continuous stream of tasks. To remember previously learned knowledge, prior studies store old samples over a memory hierarchy and replay them when new tasks arrive. Edge devices that adopt CL to preserve data privacy are typically energy-sensitive and thus require high model accuracy while not compromising energy efficiency, i.e., cost-effectiveness. Our work is the first to explore the design space of hierarchical memory replay-based CL to gain insights into achieving cost-effectiveness on edge devices. We present Miro, a novel system runtime that carefully integrates our insights into the CL framework by enabling it to dynamically configure the CL system based on resource states for the best cost-effectiveness. To reach this goal, Miro also performs online profiling on parameters with clear accuracy-energy trade-offs and adapts to optimal values with low overhead. Extensive evaluations show that Miro significantly outperforms baseline systems we build for comparison, consistently achieving higher cost-effectiveness.
Authors: Thomas Grandits, Jan Verhülsdonk, Gundolf Haase, Alexander Effland, Simone Pezzuto
The eikonal equation has become an indispensable tool for modeling cardiac electrical activation accurately and efficiently. In principle, by matching clinically recorded and eikonal-based electrocardiograms (ECGs), it is possible to build patient-specific models of cardiac electrophysiology in a purely non-invasive manner. Nonetheless, the fitting procedure remains a challenging task. The present study introduces a novel method, Geodesic-BP, to solve the inverse eikonal problem. Geodesic-BP is well-suited for GPU-accelerated machine learning frameworks, allowing us to optimize the parameters of the eikonal equation to reproduce a given ECG. We show that Geodesic-BP can reconstruct a simulated cardiac activation with high accuracy in a synthetic test case, even in the presence of modeling inaccuracies. Furthermore, we apply our algorithm to a publicly available dataset of a biventricular rabbit model, with promising results. Given the future shift towards personalized medicine, Geodesic-BP has the potential to help in future functionalizations of cardiac models meeting clinical time constraints while maintaining the physiological accuracy of state-of-the-art cardiac models.
Authors: Guillermo Bernárdez, Lev Telyatnikov, Eduard Alarcón, Albert Cabellos-Aparicio, Pere Barlet-Ros, Pietro Liò
Recently emerged Topological Deep Learning (TDL) methods aim to extend current Graph Neural Networks (GNN) by naturally processing higher-order interactions, going beyond the pairwise relations and local neighborhoods defined by graph representations. In this paper we propose a novel TDL-based method for compressing signals over graphs, consisting in two main steps: first, disjoint sets of higher-order structures are inferred based on the original signal --by clustering $N$ datapoints into $K\ll N$ collections; then, a topological-inspired message passing gets a compressed representation of the signal within those multi-element sets. Our results show that our framework improves both standard GNN and feed-forward architectures in compressing temporal link-based signals from two real-word Internet Service Provider Networks' datasets --from $30\%$ up to $90\%$ better reconstruction errors across all evaluation scenarios--, suggesting that it better captures and exploits spatial and temporal correlations over the whole graph-based network structure.
Authors: Dipesh Gyawali
Deep Learning(DL) and Machine Learning(ML) applications are rapidly increasing in recent days. Massive amounts of data are being generated over the internet which can derive meaningful results by the use of ML and DL algorithms. Hardware resources and open-source libraries have made it easy to implement these algorithms. Tensorflow and Pytorch are one of the leading frameworks for implementing ML projects. By using those frameworks, we can trace the operations executed on both GPU and CPU to analyze the resource allocations and consumption. This paper presents the time and memory allocation of CPU and GPU while training deep neural networks using Pytorch. This paper analysis shows that GPU has a lower running time as compared to CPU for deep neural networks. For a simpler network, there are not many significant improvements in GPU over the CPU.
Authors: Yuexiang Zhai, Shengbang Tong, Xiao Li, Mu Cai, Qing Qu, Yong Jae Lee, Yi Ma
Following the success of GPT4, there has been a surge in interest in multimodal large language model (MLLM) research. This line of research focuses on developing general-purpose LLMs through fine-tuning pre-trained LLMs and vision models. However, catastrophic forgetting, a notorious phenomenon where the fine-tuned model fails to retain similar performance compared to the pre-trained model, still remains an inherent problem in multimodal LLMs (MLLM). In this paper, we introduce EMT: Evaluating MulTimodality for evaluating the catastrophic forgetting in MLLMs, by treating each MLLM as an image classifier. We first apply EMT to evaluate several open-source fine-tuned MLLMs and we discover that almost all evaluated MLLMs fail to retain the same performance levels as their vision encoders on standard image classification tasks. Moreover, we continue fine-tuning LLaVA, an MLLM and utilize EMT to assess performance throughout the fine-tuning. Interestingly, our results suggest that early-stage fine-tuning on an image dataset improves performance across other image datasets, by enhancing the alignment of text and visual features. However, as fine-tuning proceeds, the MLLMs begin to hallucinate, resulting in a significant loss of generalizability, even when the image encoder remains frozen. Our results suggest that MLLMs have yet to demonstrate performance on par with their vision models on standard image classification tasks and the current MLLM fine-tuning procedure still has room for improvement.
Authors: Yukang Chen, Shengju Qian, Haotian Tang, Xin Lai, Zhijian Liu, Song Han, Jiaya Jia
We present LongLoRA, an efficient fine-tuning approach that extends the context sizes of pre-trained large language models (LLMs), with limited computation cost. Typically, training LLMs with long context sizes is computationally expensive, requiring extensive training hours and GPU resources. For example, training on the context length of 8192 needs 16x computational costs in self-attention layers as that of 2048. In this paper, we speed up the context extension of LLMs in two aspects. On the one hand, although dense global attention is needed during inference, fine-tuning the model can be effectively and efficiently done by sparse local attention. The proposed shifted sparse attention (S$^2$-Attn) effectively enables context extension, leading to non-trivial computation saving with similar performance to fine-tuning with vanilla attention. Particularly, it can be implemented with only two lines of code in training, while being optional in inference. On the other hand, we revisit the parameter-efficient fine-tuning regime for context expansion. Notably, we find that LoRA for context extension works well under the premise of trainable embedding and normalization. LongLoRA combines this improved LoRA with S$^2$-Attn. LongLoRA demonstrates strong empirical results on various tasks on Llama2 models from 7B/13B to 70B. LongLoRA adopts Llama2 7B from 4k context to 100k, or Llama2 70B to 32k on a single 8x A100 machine. LongLoRA extends models' context while retaining their original architectures, and is compatible with most existing techniques, like Flash-Attention2. In addition, we further conduct supervised fine-tuning with LongLoRA and our long instruction-following LongAlpaca dataset.
Authors: Jill-Jênn Vie (SODA), Hisashi Kashima
Knowledge tracing consists in predicting the performance of some students on new questions given their performance on previous questions, and can be a prior step to optimizing assessment and learning. Deep knowledge tracing (DKT) is a competitive model for knowledge tracing relying on recurrent neural networks, even if some simpler models may match its performance. However, little is known about why DKT works so well. In this paper, we frame deep knowledge tracing as a encoderdecoder architecture. This viewpoint not only allows us to propose better models in terms of performance, simplicity or expressivity but also opens up promising avenues for future research directions. In particular, we show on several small and large datasets that a simpler decoder, with possibly fewer parameters than the one used by DKT, can predict student performance better.
Authors: Haoyi Niu, Qimao Chen, Yingyue Li, Yi Zhang, Jianming Hu
The deployment of autonomous vehicles (AVs) has faced hurdles due to the dominance of rare but critical corner cases within the long-tail distribution of driving scenarios, which negatively affects their overall performance. To address this challenge, adversarial generation methods have emerged as a class of efficient approaches to synthesize safety-critical scenarios for AV testing. However, these generated scenarios are often underutilized for AV training, resulting in the potential for continual AV policy improvement remaining untapped, along with a deficiency in the closed-loop design needed to achieve it. Therefore, we tailor the Stackelberg Driver Model (SDM) to accurately characterize the hierarchical nature of vehicle interaction dynamics, facilitating iterative improvement by engaging background vehicles (BVs) and AV in a sequential game-like interaction paradigm. With AV acting as the leader and BVs as followers, this leader-follower modeling ensures that AV would consistently refine its policy, always taking into account the additional information that BVs play the best response to challenge AV. Extensive experiments have shown that our algorithm exhibits superior performance compared to several baselines especially in higher dimensional scenarios, leading to substantial advancements in AV capabilities while continually generating progressively challenging scenarios. Code is available at https://github.com/BlueCat-de/SDM.
Authors: Oscar Skean, Aayush Dhakal, Nathan Jacobs, Luis Gonzalo Sanchez Giraldo
Self-supervised learning (SSL) is an increasingly popular paradigm for representation learning. Recent methods can be classified as sample-contrastive, dimension-contrastive, or asymmetric network-based, with each family having its own approach to avoiding informational collapse. While dimension-contrastive methods converge to similar solutions as sample-contrastive methods, it can be empirically shown that some methods require more epochs of training to converge. Motivated by closing this divide, we present the objective function FroSSL which is both sample- and dimension-contrastive up to embedding normalization. FroSSL works by minimizing covariance Frobenius norms for avoiding collapse and minimizing mean-squared error for augmentation invariance. We show that FroSSL converges more quickly than a variety of other SSL methods and provide theoretical and empirical support that this faster convergence is due to how FroSSL affects the eigenvalues of the embedding covariance matrices. We also show that FroSSL learns competitive representations on linear probe evaluation when used to train a ResNet18 on the CIFAR-10, CIFAR-100, STL-10, and ImageNet datasets.
Authors: Zihao Wang, Yongqiang Chen, Yang Duan, Weijiang Li, Bo Han, James Cheng, Hanghang Tong
Machine Learning (ML) techniques have found applications in estimating chemical kinetic properties. With the accumulated drug molecules identified through "AI4drug discovery", the next imperative lies in AI-driven design for high-throughput chemical synthesis processes, with the estimation of properties of unseen reactions with unexplored molecules. To this end, the existing ML approaches for kinetics property prediction are required to be Out-Of-Distribution (OOD) generalizable. In this paper, we categorize the OOD kinetic property prediction into three levels (structure, condition, and mechanism), revealing unique aspects of such problems. Under this framework, we create comprehensive datasets to benchmark (1) the state-of-the-art ML approaches for reaction prediction in the OOD setting and (2) the state-of-the-art graph OOD methods in kinetics property prediction problems. Our results demonstrated the challenges and opportunities in OOD kinetics property prediction. Our datasets and benchmarks can further support research in this direction.
Authors: Andy Zhou, Kai Yan, Michal Shlapentokh-Rothman, Haohan Wang, Yu-Xiong Wang
While large language models (LLMs) have demonstrated impressive performance on a range of decision-making tasks, they rely on simple acting processes and fall short of broad deployment as autonomous agents. We introduce LATS (Language Agent Tree Search), a general framework that synergizes the capabilities of LLMs in planning, acting, and reasoning. Drawing inspiration from Monte Carlo tree search in model-based reinforcement learning, LATS employs LLMs as agents, value functions, and optimizers, repurposing their latent strengths for enhanced decision-making. What is crucial in this method is the use of an environment for external feedback, which offers a more deliberate and adaptive problem-solving mechanism that moves beyond the limitations of existing techniques. Our experimental evaluation across diverse domains, such as programming, HotPotQA, and WebShop, illustrates the applicability of LATS for both reasoning and acting. In particular, LATS achieves 94.4% for programming on HumanEval with GPT-4 and an average score of 75.9 for web browsing on WebShop with GPT-3.5, demonstrating the effectiveness and generality of our method.
Authors: Gerald Woo, Chenghao Liu, Akshat Kumar, Doyen Sahoo
Time series has been left behind in the era of pre-training and transfer learning. While research in the fields of natural language processing and computer vision are enjoying progressively larger datasets to train massive models, the most popular time series datasets consist of only tens of thousands of time steps, limiting our ability to study the effectiveness of pre-training and scaling. Recent studies have also cast doubt on the need for expressive models and scale. To alleviate these issues, we introduce three large-scale time series forecasting datasets from the cloud operations (CloudOps) domain, the largest having billions of observations, enabling further study into pre-training and scaling of time series models. We build the empirical groundwork for studying pre-training and scaling of time series models and pave the way for future research by identifying a promising candidate architecture. We show that it is a strong zero-shot baseline and benefits from further scaling, both in model and dataset size. Accompanying these datasets and results is a suite of comprehensive benchmark results comparing classical and deep learning baselines to our pre-trained method - achieving a 27% reduction in error on the largest dataset. Code and datasets can be found https://github.com/SalesforceAIResearch/pretrain-time-series-cloudops.
Authors: Simone Manti, Alessandro Lucantonio
Computational modeling is a key resource to gather insight into physical systems in modern scientific research and engineering. While access to large amount of data has fueled the use of Machine Learning (ML) to recover physical models from experiments and increase the accuracy of physical simulations, purely data-driven models have limited generalization and interpretability. To overcome these limitations, we propose a framework that combines Symbolic Regression (SR) and Discrete Exterior Calculus (DEC) for the automated discovery of physical models starting from experimental data. Since these models consist of mathematical expressions, they are interpretable and amenable to analysis, and the use of a natural, general-purpose discrete mathematical language for physics favors generalization with limited input data. Importantly, DEC provides building blocks for the discrete analogue of field theories, which are beyond the state-of-the-art applications of SR to physical problems. Further, we show that DEC allows to implement a strongly-typed SR procedure that guarantees the mathematical consistency of the recovered models and reduces the search space of symbolic expressions. Finally, we prove the effectiveness of our methodology by re-discovering three models of Continuum Physics from synthetic experimental data: Poisson equation, the Euler's Elastica and the equations of Linear Elasticity. Thanks to their general-purpose nature, the methods developed in this paper may be applied to diverse contexts of physical modeling.
Authors: Jai Pal, Bryan Hong
Artificial intelligence (AI) is a powerful tool for reshaping healthcare systems. In healthcare, AI is invaluable for its capacity to manage vast amounts of data, which can lead to more accurate and speedy diagnoses, ultimately easing the workload on healthcare professionals. As a result, AI has proven itself to be a power tool across various industries, simplifying complex tasks and pattern recognition that would otherwise be overwhelming for humans or traditional computer algorithms. In this paper, we review the strengths and weaknesses of Bayesian Ridge Regression, an AI model that can be used to bring cutting edge virus analysis to healthcare professionals around the world. The model's accuracy assessment revealed promising results, with room for improvement primarily related to data organization. In addition, the severity index serves as a valuable tool to gain a broad overview of patient care needs, aligning with healthcare professionals' preference for broader categorizations.
Authors: Seohyeon Cha, Honggu Kang, Joonhyuk Kang
Accurate uncertainty quantification in graph neural networks (GNNs) is essential, especially in high-stakes domains where GNNs are frequently employed. Conformal prediction (CP) offers a promising framework for quantifying uncertainty by providing $\textit{valid}$ prediction sets for any black-box model. CP ensures formal probabilistic guarantees that a prediction set contains a true label with a desired probability. However, the size of prediction sets, known as $\textit{inefficiency}$, is influenced by the underlying model and data generating process. On the other hand, Bayesian learning also provides a credible region based on the estimated posterior distribution, but this region is $\textit{well-calibrated}$ only when the model is correctly specified. Building on a recent work that introduced a scaling parameter for constructing valid credible regions from posterior estimate, our study explores the advantages of incorporating a temperature parameter into Bayesian GNNs within CP framework. We empirically demonstrate the existence of temperatures that result in more efficient prediction sets. Furthermore, we conduct an analysis to identify the factors contributing to inefficiency and offer valuable insights into the relationship between CP performance and model calibration.
Authors: Yufei Gu, Xiaoqing Zheng, Tomaso Aste
Double descent presents a counter-intuitive aspect within the machine learning domain, and researchers have observed its manifestation in various models and tasks. While some theoretical explanations have been proposed for this phenomenon in specific contexts, an accepted theory to account for its occurrence in deep learning remains yet to be established. In this study, we revisit the phenomenon of double descent and demonstrate that its occurrence is strongly influenced by the presence of noisy data. Through conducting a comprehensive analysis of the feature space of learned representations, we unveil that double descent arises in imperfect models trained with noisy data. We argue that double descent is a consequence of the model first learning the noisy data until interpolation and then adding implicit regularization via over-parameterization acquiring therefore capability to separate the information from the noise.
Authors: Satoshi Hayakawa, Tetsuro Morimura
Reward evaluation of episodes becomes a bottleneck in a broad range of reinforcement learning tasks. Our aim in this paper is to select a small but representative subset of a large batch of episodes, only on which we actually compute rewards for more efficient policy gradient iterations. We build a Gaussian process modeling of discounted returns or rewards to derive a positive definite kernel on the space of episodes, run an ``episodic" kernel quadrature method to compress the information of sample episodes, and pass the reduced episodes to the policy network for gradient updates. We present the theoretical background of this procedure as well as its numerical illustrations in MuJoCo tasks.
Authors: Taehyun Cho, Seungyub Han, Heesoo Lee, Kyungjae Lee, Jungwoo Lee
Distributional reinforcement learning algorithms have attempted to utilize estimated uncertainty for exploration, such as optimism in the face of uncertainty. However, using the estimated variance for optimistic exploration may cause biased data collection and hinder convergence or performance. In this paper, we present a novel distributional reinforcement learning algorithm that selects actions by randomizing risk criterion to avoid one-sided tendency on risk. We provide a perturbed distributional Bellman optimality operator by distorting the risk measure and prove the convergence and optimality of the proposed method with the weaker contraction property. Our theoretical results support that the proposed method does not fall into biased exploration and is guaranteed to converge to an optimal return. Finally, we empirically show that our method outperforms other existing distribution-based algorithms in various environments including Atari 55 games.
Authors: Shay Moran, Hilla Schefler, Jonathan Shafer
We show that many definitions of stability found in the learning theory literature are equivalent to one another. We distinguish between two families of definitions of stability: distribution-dependent and distribution-independent Bayesian stability. Within each family, we establish equivalences between various definitions, encompassing approximate differential privacy, pure differential privacy, replicability, global stability, perfect generalization, TV stability, mutual information stability, KL-divergence stability, and R\'enyi-divergence stability. Along the way, we prove boosting results that enable the amplification of the stability of a learning rule. This work is a step towards a more systematic taxonomy of stability notions in learning theory, which can promote clarity and an improved understanding of an array of stability concepts that have emerged in recent years.
Authors: Aleksandar Armacki, Pranay Sharma, Gauri Joshi, Dragana Bajovic, Dusan Jakovetic, Soummya Kar
Several recent works have studied the convergence \textit{in high probability} of stochastic gradient descent (SGD) and its clipped variant. Compared to vanilla SGD, clipped SGD is practically more stable and has the additional theoretical benefit of logarithmic dependence on the failure probability. However, the convergence of other practical nonlinear variants of SGD, e.g., sign SGD, quantized SGD and normalized SGD, that achieve improved communication efficiency or accelerated convergence is much less understood. In this work, we study the convergence bounds \textit{in high probability} of a broad class of nonlinear SGD methods. For strongly convex loss functions with Lipschitz continuous gradients, we prove a logarithmic dependence on the failure probability, even when the noise is heavy-tailed. Strictly more general than the results for clipped SGD, our results hold for any nonlinearity with bounded (component-wise or joint) outputs, such as clipping, normalization, and quantization. Further, existing results with heavy-tailed noise assume bounded $\eta$-th central moments, with $\eta \in (1,2]$. In contrast, our refined analysis works even for $\eta=1$, strictly relaxing the noise moment assumptions in the literature.
Authors: Junu Kim, Chaeeun Shim, Bosco Seong Kyu Yang, Chami Im, Sung Yoon Lim, Han-Gil Jeong, Edward Choi
Developing clinical prediction models (e.g., mortality prediction) based on electronic health records (EHRs) typically relies on expert opinion for feature selection and adjusting observation window size. This burdens experts and creates a bottleneck in the development process. We propose Retrieval-Enhanced Medical prediction model (REMed) to address such challenges. REMed can essentially evaluate an unlimited number of clinical events, select the relevant ones, and make predictions. This approach effectively eliminates the need for manual feature selection and enables an unrestricted observation window. We verified these properties through experiments on 27 clinical tasks and two independent cohorts from publicly available EHR datasets, where REMed outperformed other contemporary architectures that aim to handle as many events as possible. Notably, we found that the preferences of REMed align closely with those of medical experts. We expect our approach to significantly expedite the development of EHR prediction models by minimizing clinicians' need for manual involvement.
Authors: Mark Levy, Bruno Di Giorgi, Floris Weers, Angelos Katharopoulos, Tom Nickson
We demonstrate how conditional generation from diffusion models can be used to tackle a variety of realistic tasks in the production of music in 44.1kHz stereo audio with sampling-time guidance. The scenarios we consider include continuation, inpainting and regeneration of musical audio, the creation of smooth transitions between two different music tracks, and the transfer of desired stylistic characteristics to existing audio clips. We achieve this by applying guidance at sampling time in a simple framework that supports both reconstruction and classification losses, or any combination of the two. This approach ensures that generated audio can match its surrounding context, or conform to a class distribution or latent representation specified relative to any suitable pre-trained classifier or embedding model. Audio samples are available at https://machinelearning.apple.com/research/controllable-music
Authors: Longze Li, Jiang Chang, Aleksandar Vakanski, Yachun Wang, Tiankai Yao, Min Xian
With the increased use of data-driven approaches and machine learning-based methods in material science, the importance of reliable uncertainty quantification (UQ) of the predicted variables for informed decision-making cannot be overstated. UQ in material property prediction poses unique challenges, including the multi-scale and multi-physics nature of advanced materials, intricate interactions between numerous factors, limited availability of large curated datasets for model training, etc. Recently, Bayesian Neural Networks (BNNs) have emerged as a promising approach for UQ, offering a probabilistic framework for capturing uncertainties within neural networks. In this work, we introduce an approach for UQ within physics-informed BNNs, which integrates knowledge from governing laws in material modeling to guide the models toward physically consistent predictions. To evaluate the effectiveness of this approach, we present case studies for predicting the creep rupture life of steel alloys. Experimental validation with three datasets of collected measurements from creep tests demonstrates the ability of BNNs to produce accurate point and uncertainty estimates that are competitive or exceed the performance of the conventional method of Gaussian Process Regression. Similarly, we evaluated the suitability of BNNs for UQ in an active learning application and reported competitive performance. The most promising framework for creep life prediction is BNNs based on Markov Chain Monte Carlo approximation of the posterior distribution of network parameters, as it provided more reliable results in comparison to BNNs based on variational inference approximation or related NNs with probabilistic outputs. The codes are available at: https://github.com/avakanski/Creep-uncertainty-quantification.
Authors: Xuan Li, Zhanke Zhou, Jianing Zhu, Jiangchao Yao, Tongliang Liu, Bo Han
Despite remarkable success in various applications, large language models (LLMs) are vulnerable to adversarial jailbreaks that make the safety guardrails void. However, previous studies for jailbreaks usually resort to brute-force optimization or extrapolations of a high computation cost, which might not be practical or effective. In this paper, inspired by the Milgram experiment that individuals can harm another person if they are told to do so by an authoritative figure, we disclose a lightweight method, termed as DeepInception, which can easily hypnotize LLM to be a jailbreaker and unlock its misusing risks. Specifically, DeepInception leverages the personification ability of LLM to construct a novel nested scene to behave, which realizes an adaptive way to escape the usage control in a normal scenario and provides the possibility for further direct jailbreaks. Empirically, we conduct comprehensive experiments to show its efficacy. Our DeepInception can achieve competitive jailbreak success rates with previous counterparts and realize a continuous jailbreak in subsequent interactions, which reveals the critical weakness of self-losing on both open/closed-source LLMs like Falcon, Vicuna, Llama-2, and GPT-3.5/4/4V. Our investigation appeals that people should pay more attention to the safety aspects of LLMs and a stronger defense against their misuse risks. The code is publicly available at: https://github.com/tmlr-group/DeepInception.
Authors: Haotian Zheng, Qizhou Wang, Zhen Fang, Xiaobo Xia, Feng Liu, Tongliang Liu, Bo Han
Out-of-distribution (OOD) detection discerns OOD data where the predictor cannot make valid predictions as in-distribution (ID) data, thereby increasing the reliability of open-world classification. However, it is typically hard to collect real out-of-distribution (OOD) data for training a predictor capable of discerning ID and OOD patterns. This obstacle gives rise to data generation-based learning methods, synthesizing OOD data via data generators for predictor training without requiring any real OOD data. Related methods typically pre-train a generator on ID data and adopt various selection procedures to find those data likely to be the OOD cases. However, generated data may still coincide with ID semantics, i.e., mistaken OOD generation remains, confusing the predictor between ID and OOD data. To this end, we suggest that generated data (with mistaken OOD generation) can be used to devise an auxiliary OOD detection task to facilitate real OOD detection. Specifically, we can ensure that learning from such an auxiliary task is beneficial if the ID and the OOD parts have disjoint supports, with the help of a well-designed training procedure for the predictor. Accordingly, we propose a powerful data generation-based learning method named Auxiliary Task-based OOD Learning (ATOL) that can relieve the mistaken OOD generation. We conduct extensive experiments under various OOD detection setups, demonstrating the effectiveness of our method against its advanced counterparts.
Authors: Colin Troisemaine, Alexandre Reiffers-Masson, Stéphane Gosselin, Vincent Lemaire, Sandrine Vaton
The problem of Novel Class Discovery (NCD) consists in extracting knowledge from a labeled set of known classes to accurately partition an unlabeled set of novel classes. While NCD has recently received a lot of attention from the community, it is often solved on computer vision problems and under unrealistic conditions. In particular, the number of novel classes is usually assumed to be known in advance, and their labels are sometimes used to tune hyperparameters. Methods that rely on these assumptions are not applicable in real-world scenarios. In this work, we focus on solving NCD in tabular data when no prior knowledge of the novel classes is available. To this end, we propose to tune the hyperparameters of NCD methods by adapting the $k$-fold cross-validation process and hiding some of the known classes in each fold. Since we have found that methods with too many hyperparameters are likely to overfit these hidden classes, we define a simple deep NCD model. This method is composed of only the essential elements necessary for the NCD problem and performs impressively well under realistic conditions. Furthermore, we find that the latent space of this method can be used to reliably estimate the number of novel classes. Additionally, we adapt two unsupervised clustering algorithms ($k$-means and Spectral Clustering) to leverage the knowledge of the known classes. Extensive experiments are conducted on 7 tabular datasets and demonstrate the effectiveness of the proposed method and hyperparameter tuning process, and show that the NCD problem can be solved without relying on knowledge from the novel classes.
Authors: Joao P. A. Dantas, Diego Geraldo, Felipe L. L. Medeiros, Marcos R. O. A. Maximo, Takashi Yoneyama
Surface-to-Air Missiles (SAMs) are crucial in modern air defense systems. A critical aspect of their effectiveness is the Engagement Zone (EZ), the spatial region within which a SAM can effectively engage and neutralize a target. Notably, the EZ is intrinsically related to the missile's maximum range; it defines the furthest distance at which a missile can intercept a target. The accurate computation of this EZ is essential but challenging due to the dynamic and complex factors involved, which often lead to high computational costs and extended processing times when using conventional simulation methods. In light of these challenges, our study investigates the potential of machine learning techniques, proposing an approach that integrates machine learning with a custom-designed simulation tool to train supervised algorithms. We leverage a comprehensive dataset of pre-computed SAM EZ simulations, enabling our model to accurately predict the SAM EZ for new input parameters. It accelerates SAM EZ simulations, enhances air defense strategic planning, and provides real-time insights, improving SAM system performance. The study also includes a comparative analysis of machine learning algorithms, illuminating their capabilities and performance metrics and suggesting areas for future research, highlighting the transformative potential of machine learning in SAM EZ simulations.
Authors: Sai Akash Bathini, Dagli Cihan
The application of Machine learning to finance has become a familiar approach, even more so in stock market forecasting. The stock market is highly volatile and huge amounts of data are generated every minute globally. The extraction of effective intelligence from this data is of critical importance. However, a collaboration of numerical stock data with qualitative text data can be a challenging task. In this work, we accomplish this and provide an unprecedented, publicly available dataset with technical and fundamental data, sentiment that we gathered from News Archives, TV news captions, Radio Transcripts, Tweets, Daily financial newspapers, etc. The text data entries used for sentiment extraction total more than 1.4 Million. The dataset consists of daily entries from January 2018 to December 2022 for 8 companies representing diverse industrial sectors and the Dow Jones Industrial Average (DJIA) as a whole. Holistic Fundamental and Technical data is provided training ready for Model learning and deployment. The data generated could be used for Incremental online learning with real-time data points retrieved daily, since there was no stagnant data utilized, all the data was retired from APIs or self-designed scripts. Moreover, the utilization of Spearman's rank correlation over real-time data, linking stock returns with sentiment analysis has produced noteworthy results for the DJIA achieving accuracy levels surpassing 60\%. The dataset is made available at https://github.com/batking24/Huge-Stock-Dataset
Authors: Yan Kang, Tao Fan, Hanlin Gu, Lixin Fan, Qiang Yang
Foundation Models (FMs) such as GPT-4 encoded with vast knowledge and powerful emergent abilities have achieved remarkable success in various natural language processing and computer vision tasks. Grounding FMs by adapting them to domain-specific tasks or augmenting them with domain-specific knowledge enables us to exploit the full potential of FMs. However, grounding FMs faces several challenges, stemming primarily from constrained computing resources, data privacy, model heterogeneity, and model ownership. Federated Transfer Learning (FTL), the combination of federated learning and transfer learning, provides promising solutions to address these challenges. In recent years, the need for grounding FMs leveraging FTL, coined FTL-FM, has arisen strongly in both academia and industry. Motivated by the strong growth in FTL-FM research and the potential impact of FTL-FM on industrial applications, we propose an FTL-FM framework that formulates problems of grounding FMs in the federated learning setting, construct a detailed taxonomy based on the FTL-FM framework to categorize state-of-the-art FTL-FM works, and comprehensively overview FTL-FM works based on the proposed taxonomy. We also establish correspondences between FTL-FM and conventional phases of adapting FM so that FM practitioners can align their research works with FTL-FM. In addition, we overview advanced efficiency-improving and privacy-preserving techniques because efficiency and privacy are critical concerns in FTL-FM. Last, we discuss opportunities and future research directions of FTL-FM.
Authors: Yuyang Wang, Ahmed A. Elhag, Navdeep Jaitly, Joshua M. Susskind, Miguel Angel Bautista
In this paper we tackle the problem of generating conformers of a molecule in 3D space given its molecular graph. We parameterize these conformers as continuous functions that map elements from the molecular graph to points in 3D space. We then formulate the problem of learning to generate conformers as learning a distribution over these functions using a diffusion generative model, called Molecular Conformer Fields (MCF). Our approach is simple and scalable, and achieves state-of-the-art performance on challenging molecular conformer generation benchmarks while making no assumptions about the explicit structure of molecules (e.g. modeling torsional angles). MCF represents an advance in extending diffusion models to handle complex scientific problems in a conceptually simple, scalable and effective manner.
Authors: Xiao Liu, Xuanyu Lei, Shengyuan Wang, Yue Huang, Zhuoer Feng, Bosi Wen, Jiale Cheng, Pei Ke, Yifan Xu, Weng Lam Tam, Xiaohan Zhang, Lichao Sun, Hongning Wang, Jing Zhang, Minlie Huang, Yuxiao Dong, Jie Tang
Alignment has become a critical step for instruction-tuned Large Language Models (LLMs) to become helpful assistants. However, effective evaluation of alignment for emerging Chinese LLMs is still significantly lacking, calling for real-scenario grounded, open-ended, challenging and automatic evaluations tailored for alignment. To fill in this gap, we introduce AlignBench, a comprehensive multi-dimensional benchmark for evaluating LLMs' alignment in Chinese. Equipped with a human-in-the-loop data curation pipeline, our benchmark employs a rule-calibrated multi-dimensional LLM-as-Judge with Chain-of-Thought to generate explanations and final ratings as evaluations, ensuring high reliability and interpretability. Furthermore, we report AlignBench evaluated by CritiqueLLM, a dedicated Chinese evaluator LLM that recovers 95% of GPT-4's evaluation ability. We will provide public APIs for evaluating AlignBench with CritiqueLLM to facilitate the evaluation of LLMs' Chinese alignment. All evaluation codes, data, and LLM generations are available at \url{https://github.com/THUDM/AlignBench}.
Authors: Kaiwen Hou
This manuscript enriches the framework of continuous normalizing flows (CNFs) within causal inference, primarily to augment the geometric properties of parametric submodels used in targeted maximum likelihood estimation (TMLE). By introducing an innovative application of CNFs, we construct a refined series of parametric submodels that enable a directed interpolation between the prior distribution $p_0$ and the empirical distribution $p_1$. This proposed methodology serves to optimize the semiparametric efficiency bound in causal inference by orchestrating CNFs to align with Wasserstein gradient flows. Our approach not only endeavors to minimize the mean squared error in the estimation but also imbues the estimators with geometric sophistication, thereby enhancing robustness against misspecification. This robustness is crucial, as it alleviates the dependence on the standard $n^{\frac{1}{4}}$ rate for a doubly-robust perturbation direction in TMLE. By incorporating robust optimization principles and differential geometry into the estimators, the developed geometry-aware CNFs represent a significant advancement in the pursuit of doubly robust causal inference.
Authors: Lei Guan, Dongsheng Li, Jiye Liang, Wenjian Wang, Xicheng Lu
Asynchronous pipeline model parallelism with a "1F1B" (one forward, one backward) schedule generates little bubble overhead and always provides quite a high throughput. However, the "1F1B" schedule inevitably leads to weight inconsistency and weight staleness issues due to the cross-training of different mini-batches across GPUs. To simultaneously address these two problems, in this paper, we propose an optimizer-dependent weight prediction strategy (a.k.a PipeOptim) for asynchronous pipeline training. The key insight of our proposal is that we employ a weight prediction strategy in the forward pass to ensure that each mini-batch uses consistent and staleness-free weights to compute the forward pass. To be concrete, we first construct the weight prediction scheme based on the update rule of the used optimizer when training the deep neural network models. Then throughout the "1F1B" pipelined training, each mini-batch is mandated to execute weight prediction ahead of the forward pass, subsequently employing the predicted weights to perform the forward pass. As a result, PipeOptim 1) inherits the advantage of the "1F1B" schedule and generates pretty high throughput, and 2) can ensure effective parameter learning regardless of the type of the used optimizer. To verify the effectiveness of our proposal, we conducted extensive experimental evaluations using eight different deep-learning models spanning three machine-learning tasks including image classification, sentiment analysis, and machine translation. The experiment results demonstrate that PipeOptim outperforms the popular pipelined approaches including GPipe, PipeDream, PipeDream-2BW, and SpecTrain. The code of PipeOptim can be accessible at https://github.com/guanleics/PipeOptim.
Authors: Rémi Munos, Michal Valko, Daniele Calandriello, Mohammad Gheshlaghi Azar, Mark Rowland, Zhaohan Daniel Guo, Yunhao Tang, Matthieu Geist, Thomas Mesnard, Andrea Michi, Marco Selvi, Sertan Girgin, Nikola Momchev, Olivier Bachem, Daniel J. Mankowitz, Doina Precup, Bilal Piot
Reinforcement learning from human feedback (RLHF) has emerged as the main paradigm for aligning large language models (LLMs) with human preferences. Typically, RLHF involves the initial step of learning a reward model from human feedback, often expressed as preferences between pairs of text generations produced by a pre-trained LLM. Subsequently, the LLM's policy is fine-tuned by optimizing it to maximize the reward model through a reinforcement learning algorithm. However, an inherent limitation of current reward models is their inability to fully represent the richness of human preferences and their dependency on the sampling distribution.
In this study, we introduce an alternative pipeline for the fine-tuning of LLMs using pairwise human feedback. Our approach entails the initial learning of a preference model, which is conditioned on two inputs given a prompt, followed by the pursuit of a policy that consistently generates responses preferred over those generated by any competing policy, thus defining the Nash equilibrium of this preference model. We term this approach Nash learning from human feedback (NLHF).
In the context of a tabular policy representation, we present a novel algorithmic solution, Nash-MD, founded on the principles of mirror descent. This algorithm produces a sequence of policies, with the last iteration converging to the regularized Nash equilibrium. Additionally, we explore parametric representations of policies and introduce gradient descent algorithms for deep-learning architectures. To demonstrate the effectiveness of our approach, we present experimental results involving the fine-tuning of a LLM for a text summarization task. We believe NLHF offers a compelling avenue for preference learning and policy optimization with the potential of advancing the field of aligning LLMs with human preferences.
Authors: Edan Meyer, Adam White, Marlos C. Machado
Reinforcement learning (RL) agents make decisions using nothing but observations from the environment, and consequently, heavily rely on the representations of those observations. Though some recent breakthroughs have used vector-based categorical representations of observations, often referred to as discrete representations, there is little work explicitly assessing the significance of such a choice. In this work, we provide a thorough empirical investigation of the advantages of representing observations as vectors of categorical values within the context of reinforcement learning. We perform evaluations on world-model learning, model-free RL, and ultimately continual RL problems, where the benefits best align with the needs of the problem setting. We find that, when compared to traditional continuous representations, world models learned over discrete representations accurately model more of the world with less capacity, and that agents trained with discrete representations learn better policies with less data. In the context of continual RL, these benefits translate into faster adapting agents. Additionally, our analysis suggests that the observed performance improvements can be attributed to the information contained within the latent vectors and potentially the encoding of the discrete representation itself.
Authors: Raghavv Goel, Cecilia Morales, Manpreet Singh, Artur Dubrawski, John Galeotti, Howie Choset
Segmenting a moving needle in ultrasound images is challenging due to the presence of artifacts, noise, and needle occlusion. This task becomes even more demanding in scenarios where data availability is limited. Convolutional Neural Networks (CNNs) have been successful in many computer vision applications, but struggle to accurately segment needles without considering their motion. In this paper, we present a novel approach for needle segmentation that combines classical Kalman Filter (KF) techniques with data-driven learning, incorporating both needle features and needle motion. Our method offers two key contributions. First, we propose a compatible framework that seamlessly integrates into commonly used encoder-decoder style architectures. Second, we demonstrate superior performance compared to recent state-of-the-art needle segmentation models using our novel convolutional neural network (CNN) based KF-inspired block, achieving a 15\% reduction in pixel-wise needle tip error and an 8\% reduction in length error. Third, to our knowledge we are the first to implement a learnable filter to incorporate non-linear needle motion for improving needle segmentation.
Authors: Che Liu, Cheng Ouyang, Yinda Chen, Cesar César Quilodrán-Casas, Lei Ma, Jie Fu, Yike Guo, Anand Shah, Wenjia Bai, Rossella Arcucci
Expert annotation of 3D medical image for downstream analysis is resource-intensive, posing challenges in clinical applications. Visual self-supervised learning (vSSL), though effective for learning visual invariance, neglects the incorporation of domain knowledge from medicine. To incorporate medical knowledge into visual representation learning, vision-language pre-training (VLP) has shown promising results in 2D image. However, existing VLP approaches become generally impractical when applied to high-resolution 3D medical images due to GPU hardware constraints and the potential loss of critical details caused by downsampling, which is the intuitive solution to hardware constraints. To address the above limitations, we introduce T3D, the first VLP framework designed for high-resolution 3D medical images. T3D incorporates two text-informed pretext tasks: (\lowerromannumeral{1}) text-informed contrastive learning; (\lowerromannumeral{2}) text-informed image restoration. These tasks focus on learning 3D visual representations from high-resolution 3D medical images and integrating clinical knowledge from radiology reports, without distorting information through forced alignment of downsampled volumes with detailed anatomical text. Trained on a newly curated large-scale dataset of 3D medical images and radiology reports, T3D significantly outperforms current vSSL methods in tasks like organ and tumor segmentation, as well as disease classification. This underlines T3D's potential in representation learning for 3D medical image analysis. All data and code will be available upon acceptance.
Authors: Amena Darwish, Stefan Ericson, Rohollah Ghasemi, Tobias Andersson, Dan Lönn, Andreas Andersson Lassila, Kent Salomonsson
To advance quality assurance in the welding process, this study presents a robust deep learning model that enables the prediction of two critical welds Key Performance Characteristics (KPCs): welding depth and average pore volume. In the proposed approach, a comprehensive range of laser welding Key Input Characteristics (KICs) is utilized, including welding beam geometries, welding feed rates, path repetitions for weld beam geometries, and bright light weld ratios for all paths, all of which were obtained from hairpin welding experiments. Two deep learning networks are employed with multiple hidden dense layers and linear activation functions to showcase the capabilities of deep neural networks in capturing the intricate nonlinear connections inherent within welding KPCs and KICs. Applying deep learning networks to the small numerical experimental hairpin welding dataset has shown promising results, achieving Mean Absolute Error (MAE) values as low as 0.1079 for predicting welding depth and 0.0641 for average pore volume. Additionally, the validity verification demonstrates the reliability of the proposed method. This, in turn, promises significant advantages in controlling welding outcomes, moving beyond the current trend of relying merely on monitoring for defect classification.
Authors: Haochen Zhang, Yuyang Dong, Chuan Xiao, Masafumi Oyamada
In this paper, we present Jellyfish, an open-source LLM as a universal task solver for DP. Built on the Llama 2 13B model, Jellyfish is instruction-tuned with the datasets of several typical DP tasks including error detection, data imputation, schema matching, and entity matching, and delivers generalizability to other tasks. Remarkably, Jellyfish can operate on a local, single, and low-priced GPU with its 13 billion parameters, ensuring data security and enabling further tuning. Its proficiency in understanding natural language allows users to manually craft instructions for DP tasks. Unlike many existing methods that heavily rely on prior knowledge, Jellyfish acquires domain knowledge during its tuning process and integrates optional knowledge injection during inference. A distinctive feature of Jellyfish is its interpreter, which elucidates its output decisions. To construct Jellyfish, we develop a series of pre-tuning and DP-tuning techniques. Jellyfish is equipped with an instance serializer, which automatically translates raw data into model prompts, and a knowledge injector, which optionally introduces task- and dataset-specific knowledge to enhance DP performance. Our evaluation of Jellyfish, using a range of real datasets, shows its competitiveness compared to state-of-the-art methods and its strong generalizability to unseen tasks. Jellyfish's performance rivals that of GPT series models, and its interpreter offers enhanced reasoning capabilities compared to GPT-3.5. Furthermore, our evaluation highlights the effectiveness of the techniques employed in constructing Jellyfish. Our model is available at Hugging Face: https://huggingface.co/NECOUDBFM/Jellyfish .
Authors: Lucas Farndale, Robert Insall, Ke Yuan
Computational pathology models rarely utilise data that will not be available for inference. This means most models cannot learn from highly informative data such as additional immunohistochemical (IHC) stains and spatial transcriptomics. We present TriDeNT, a novel self-supervised method for utilising privileged data that is not available during inference to improve performance. We demonstrate the efficacy of this method for a range of different paired data including immunohistochemistry, spatial transcriptomics and expert nuclei annotations. In all settings, TriDeNT outperforms other state-of-the-art methods in downstream tasks, with observed improvements of up to 101%. Furthermore, we provide qualitative and quantitative measurements of the features learned by these models and how they differ from baselines. TriDeNT offers a novel method to distil knowledge from scarce or costly data during training, to create significantly better models for routine inputs.
Authors: Trong Duong, Sang T. Truong, Minh Tam, Bao Bach, Ju-Young Ryu, June-Koo Kevin Rhee
Quantum neural networks are promising for a wide range of applications in the Noisy Intermediate-Scale Quantum era. As such, there is an increasing demand for automatic quantum neural architecture search. We tackle this challenge by designing a quantum circuits metric for Bayesian optimization with Gaussian process. To this goal, we propose a new quantum gates distance that characterizes the gates' action over every quantum state and provide a theoretical perspective on its geometrical properties. Our approach significantly outperforms the benchmark on three empirical quantum machine learning problems including training a quantum generative adversarial network, solving combinatorial optimization in the MaxCut problem, and simulating quantum Fourier transform. Our method can be extended to characterize behaviors of various quantum machine learning models.
Authors: Nick Whiteley, Annie Gray, Patrick Rubin-Delanchy
The Manifold Hypothesis is a widely accepted tenet of Machine Learning which asserts that nominally high-dimensional data are in fact concentrated near a low-dimensional manifold, embedded in high-dimensional space. This phenomenon is observed empirically in many real world situations, has led to development of a wide range of statistical methods in the last few decades, and has been suggested as a key factor in the success of modern AI technologies. We show that rich and sometimes intricate manifold structure in data can emerge from a generic and remarkably simple statistical model -- the Latent Metric Model -- via elementary concepts such as latent variables, correlation and stationarity. This establishes a general statistical explanation for why the Manifold Hypothesis seems to hold in so many situations. Informed by the Latent Metric Model we derive procedures to discover and interpret the geometry of high-dimensional data, and explore hypotheses about the data generating mechanism. These procedures operate under minimal assumptions and make use of well known, scaleable graph-analytic algorithms.