论文检索

Theory · Learning Theory

Chi-Heng Lin, Chiraag Kaushik, Eva Dyer, Vidya Muthukumar

Data augmentation (DA) is a powerful workhorse for bolstering performance in modern machine learning. Specific augmentations like translations and scaling in computer vision are traditionally believed to improve generalization by generating new (artificial) data from the same distribution. However, this traditional viewpoint does not explain the success of prevalent augmentations in modern machine learning (e.g. randomized masking, cutout, mixup), that greatly alter the training data distribution. In this work, we develop a new theoretical framework to characterize the impact of a general class of DA on underparameterized and overparameterized linear model generalization. Our framework reveals that DA induces implicit spectral regularization through a combination of two distinct effects: a) manipulating the relative proportion of eigenvalues of the data covariance matrix in a training-data-dependent manner, and b) uniformly boosting the entire spectrum of the data covariance matrix through ridge regression. These effects, when applied to popular augmentations, give rise to a wide variety of phenomena, including discrepancies in generalization between overparameterized and underparameterized regimes and differences between regression and classification tasks. Our framework highlights the nuanced and sometimes surprising impacts of DA on generalization, and serves as a testbed for novel augmentation design.

Deep Learning · Other Representation Learning

Chiraag Kaushik, Ran Liu, Chi-Heng Lin, Amrit Khera, Matthew Jin, Wenrui Ma, Vidya Muthukumar, Eva Dyer

Classification models are expected to perform equally well for different classes, yet in practice, there are often large gaps in their performance. This issue of class bias is widely studied in cases of datasets with sample imbalance, but is relatively overlooked in balanced datasets. In this work, we introduce the concept of spectral imbalance in features as a potential source for class disparities and study the connections between spectral imbalance and class bias in both theory and practice. To build the connection between spectral imbalance and class gap, we develop a theoretical framework for studying class disparities and derive exact expressions for the per-class error in a high-dimensional mixture model setting. We then study this phenomenon in 11 different state-of-the-art pre-trained encoders, and show how our proposed framework can be used to compare the quality of encoders, as well as evaluate and combine data augmentation strategies to mitigate the issue. Our work sheds light on the class-dependent effects of learning, and provides new insights into how state-of-the-art pre-trained features may have unknown biases that can be diagnosed through their spectra.

Reinforcement Learning

Chengjie Wu, Hao Hu, yiqin yang, Ning Zhang, Chongjie Zhang

The surge in volumes of video data offers unprecedented opportunities for advancing reinforcement learning (RL). This growth has motivated the development of passive RL, seeking to convert passive observations into actionable insights. This paper explores the prerequisites and mechanisms through which passive data can be utilized to improve online RL. We show that, in identifiable dynamics, where action impact can be distinguished from stochasticity, learning on passive data is statistically beneficial. Building upon the theoretical insights, we propose a novel algorithm named Multiscale State-Centric Planners (MSCP) that leverages two planners at distinct scales to offer guidance across varying levels of abstraction. The algorithm's fast planner targets immediate objectives, while the slow planner focuses on achieving longer-term goals. Notably, the fast planner incorporates pessimistic regularization to address the distributional shift between offline and online data. MSCP effectively handles the practical challenges involving imperfect pretraining and limited dataset coverage. Our empirical evaluations across multiple benchmarks demonstrate that MSCP significantly outperforms existing approaches, underscoring its proficiency in addressing complex, long-horizon tasks through the strategic use of passive data.

Social Aspects · Privacy-preserving Statistics and Machine Learning

Jiachun Li, Kaining Shi, David Simchi-Levi

Adaptive experiment is widely adopted to estimate conditional average treatment effect (CATE) in clinical trials and many other scenarios. While the primary goal in experiment is to maximize estimation accuracy, due to the imperative of social welfare, it's also crucial to provide treatment with superior outcomes to patients, which is measured by regret in contextual bandit framework. Furthermore, privacy concerns arise in clinical scenarios containing sensitive data like patients health records. Therefore, it's essential for the treatment allocation mechanism to incorporate robust privacy protection measures. In this paper, we investigate the tradeoff between loss of social welfare and statistical power of CATE estimation in contextual bandit experiment. We propose a matched upper and lower bound for the multi-objective optimization problem, and then adopt the concept of Pareto optimality to mathematically characterize the optimality condition. Furthermore, we propose differentially private algorithms which still matches the lower bound, showing that privacy is "almost free". Additionally, we derive the asymptotic normality of the estimator, which is essential in statistical inference and hypothesis testing.

Jiawei Zhao, Zhenyu Zhang, Beidi Chen, Zhangyang “Atlas” Wang, Anima Anandkumar, Yuandong Tian

Training Large Language Models (LLMs) presents significant memory challenges, predominantly due to the growing size of weights and optimizer states. Common memory-reduction approaches, such as low-rank adaptation (LoRA), add a trainable low-rank matrix to the frozen pre-trained weight in each layer, reducing trainable parameters and optimizer states. However, such approaches typically underperform training with full-rank weights in both pre-training and fine-tuning stages since they limit the parameter search to a low-rank subspace and alter the training dynamics, and further, may require full-rank warm start. In this work, we propose Gradient Low-Rank Projection (GaLore), a training strategy that allows full-parameter learning but is more memory-efficient than common low-rank adaptation methods such as LoRA. Our approach reduces memory usage by up to 65.5% in optimizer states while maintaining both efficiency and performance for pre-training on LLaMA 1B and 7B architectures with C4 dataset with up to 19.7B tokens, and on fine-tuning RoBERTa on GLUE tasks. Our 8-bit GaLore further reduces optimizer memory by up to 82.5% and total training memory by 63.3%, compared to a BF16 baseline. Notably, we demonstrate, for the first time, the feasibility of pre-training a 7B model on consumer GPUs with 24GB memory (e.g., NVIDIA RTX 4090) without model parallel, checkpointing, or offloading strategies.

Haotong Qin, Xudong Ma, Xingyu Zheng, Xiaoyang Li, Yang Zhang, Shouda Liu, Jie Luo, Xianglong Liu, Michele Magno

The LoRA-finetuning quantization of LLMs has been extensively studied to obtain accurate yet compact LLMs for deployment on resource-constrained hardware. However, existing methods cause the quantized LLM to severely degrade and even fail to benefit from the finetuning of LoRA. This paper proposes a novel IR-QLoRA for pushing quantized LLMs with LoRA to be highly accurate through information retention. The proposed IR-QLoRA mainly relies on two technologies derived from the perspective of unified information: (1) statistics-based Information Calibration Quantization allows the quantized parameters of LLM to retain original information accurately; (2) finetuning-based Information Elastic Connection makes LoRA utilizes elastic representation transformation with diverse information. Comprehensive experiments show that IR-QLoRA can significantly improve accuracy across LLaMA and LLaMA2 families under 2-4 bit-widths, e.g., 4-bit LLaMA-7B achieves 1.4% improvement on MMLU compared with the state-of-the-art methods. The significant performance gain requires only a tiny 0.31% additional time consumption, revealing the satisfactory efficiency of our IR-QLoRA. We highlight that IR-QLoRA enjoys excellent versatility, compatible with various frameworks (e.g., NormalFloat and Integer quantization) and brings general accuracy gains. The code is available at https://github.com/htqin/ir-qlora .

Reinforcement Learning

Samuel Garcin, James Doran, Shangmin Guo, Christopher Lucas, Stefano V. Albrecht

Autonomous agents trained using deep reinforcement learning (RL) often lack the ability to successfully generalise to new environments, even when these environments share characteristics with the ones they have encountered during training. In this work, we investigate how the sampling of individual environment instances, or levels, affects the zero-shot generalisation (ZSG) ability of RL agents. We discover that, for deep actor-critic architectures sharing their base layers, prioritising levels according to their value loss minimises the mutual information between the agent's internal representation and the set of training levels in the generated training data. This provides a novel theoretical justification for the regularisation achieved by certain adaptive sampling strategies. We then turn our attention to unsupervised environment design (UED) methods, which assume control over level generation. We find that existing UED methods can significantly shift the training distribution, which translates to low ZSG performance. To prevent both overfitting and distributional shift, we introduce *data-regularised environment design* (DRED). DRED generates levels using a generative model trained to approximate the ground truth distribution of an initial set of level parameters. Through its grounding, DRED achieves significant improvements in ZSG over adaptive level sampling strategies and UED methods.

Deep Learning · Robustness

Yongshuo Zong, Tingyang Yu, Ruchika Chavhan, Bingchen Zhao, Timothy Hospedales

Large language and vision-language models are rapidly being deployed in practice thanks to their impressive capabilities in instruction following, in-context learning, and so on. This raises an urgent need to carefully analyse their robustness so that stakeholders can understand if and when such models are trustworthy enough to be relied upon in any given application. In this paper, we highlight a specific vulnerability in popular models, namely permutation sensitivity in multiple-choice question answering (MCQA). Specifically, we show empirically that popular models are vulnerable to adversarial permutation in answer sets for multiple-choice prompting, which is surprising as models should ideally be as invariant to prompt permutation as humans are. These vulnerabilities persist across various model sizes, and exist in very recent language and vision-language models. Code to reproduce all experiments is provided in supplementary materials.

Probabilistic Methods

JIAN XU, Delu Zeng, John Paisley

Deep Gaussian processes (DGPs) provide a robust paradigm in Bayesian deep learning. In DGPs, a set of sparse integration locations called inducing points are selected to approximate the posterior distribution of the model. This is done to reduce computational complexity and improve model efficiency. However, inferring the posterior distribution of inducing points is not straightforward. Traditional variational inference techniques methods to approximate the posterior often leads to significant bias. To address this issue, we propose an alternative named Denoising Diffusion Variational Inference (DDVI) that utilizes a denoising diffusion stochastic differential equation (SDE) for generating posterior samples of inducing variables. We refer to the score matching method in the denoising diffusion model to approximate challenging score functions using a neural network. Furthermore, by combining classical mathematical theory of SDE with the minimization of KL divergence between the approximate and true processes, we propose a novel explicit variational lower bound for the marginal likelihood function of DGP. Through extensive experiments on various datasets and comparisons with baseline methods, we empirically demonstrate the effectiveness of the DDVI method in posterior inference of inducing points for DGP models.

Social Aspects · Trustworthy Machine Learning

Youwei Shu, Xi Xiao, Derui Wang, Yuxin Cao, Siji Chen, Minhui Xue, Linyi Li, Bo Li

Randomized Smoothing (RS) is currently a scalable certified defense method providing robustness certification against adversarial examples. Although significant progress has been achieved in providing defenses against $\ell_p$ adversaries, the interaction between the smoothing distribution and the robustness certification still remains vague. In this work, we comprehensively study the effect of two families of distributions, named Exponential Standard Gaussian (ESG) and Exponential General Gaussian (EGG) distributions, on Randomized Smoothing and Double Sampling Randomized Smoothing (DSRS). We derive an analytic formula for ESG's certified radius, which converges to the origin formula of RS as the dimension $d$ increases. Additionally, we prove that EGG can provide tighter constant factors than DSRS in providing $\Omega(\sqrt{d})$ lower bounds of $\ell_2$ certified radius, and thus further addresses the curse of dimensionality in RS. Our experiments on real-world datasets confirm our theoretical analysis of the ESG distributions, that they provide almost the same certification under different exponents $\eta$ for both RS and DSRS. In addition, EGG brings a significant improvement to the DSRS certification, but the mechanism can be different when the classifier properties are different. Compared to the primitive DSRS, the increase in certified accuracy provided by EGG is prominent, up to 6.4% on ImageNet.

Deep Learning · Generative Models and Autoencoders

Randall Balestriero, Yann LeCun

Input space reconstruction is an attractive representation learning paradigm. Despite interpretability benefit of reconstruction and generation, we identify a misalignment between learning to reconstruct, and learning for perception. We show that the former allocates a model's capacity towards a subspace of the data explaining the observed variance--a subspace with uninformative features for the latter. For example, the supervised TinyImagenet task with images projected onto the top subspace explaining 90% of the pixel variance can be solved with 45% test accuracy. Using the bottom subspace instead, accounting for only 20% of the pixel variance, reaches 55% test accuracy. Learning by reconstruction is also wasteful as the features for perception are learned last, pushing the need for long training schedules. We finally prove that learning by denoising can alleviate that misalignment for some noise strategies, e.g., masking. While tuning the noise strategy without knowledge of the perception task seems challenging, we provide a solution to detect if a noise strategy is never beneficial regardless of the perception task, e.g., additive Gaussian noise.

Deep Learning

Jintong Gao, He Zhao, Dandan Guo, Hongyuan Zha

A well-trained deep neural network on balanced datasets usually exhibits the Neural Collapse (NC) phenomenon, which is an informative indicator of the model achieving good performance. However, NC is usually hard to be achieved for a model trained on long-tailed datasets, leading to the deteriorated performance of test data. This work aims to induce the NC phenomenon in imbalanced learning from the perspective of distribution matching. By enforcing the distribution of last-layer representations to align the ideal distribution of the ETF structure, we develop a Distribution Alignment Optimization (DisA) loss, acting as a plug-and-play method can be combined with most of the existing long-tailed methods, we further instantiate it to the cases of fixing classifier and learning classifier. The extensive experiments show the effectiveness of DisA, providing a promising solution to the imbalanced issue. Our code is available at DisA.

Deep Learning · Algorithms

Huy Tran, Yikun Bai, Abihith Kothapalli, Ashkan Shahbazi, XINRAN LIU, Rocio Diaz Martin, Soheil Kolouri

Comparing spherical probability distributions is of great interest in various fields, including geology, medical domains, computer vision, and deep representation learning. The utility of optimal transport-based distances, such as the Wasserstein distance, for comparing probability measures has spurred active research in developing computationally efficient variations of these distances for spherical probability measures. This paper introduces a high-speed and highly parallelizable distance for comparing spherical measures using the stereographic projection and the generalized Radon transform, which we refer to as the Stereographic Spherical Sliced Wasserstein (S3W) distance. We carefully address the distance distortion caused by the stereographic projection and provide an extensive theoretical analysis of our proposed metric and its rotationally invariant variation. Finally, we evaluate the performance of the proposed metrics and compare them with recent baselines in terms of both speed and accuracy through a wide range of numerical studies, including gradient flows and self-supervised learning. Our code is available at https://github.com/mint-vu/s3wd.

Miscellaneous Aspects of Machine Learning · Causality

Junyi Zou, Matthew Levine, Dessi Zaharieva, Ramesh Johari, Emily Fox

Hybrid models composing mechanistic ODE-based dynamics with flexible and expressive neural network components have grown rapidly in popularity, especially in scientific domains where such ODE-based modeling offers important interpretability and validated causal grounding (e.g., for counterfactual reasoning). The incorporation of mechanistic models also provides inductive bias in standard blackbox modeling approaches, critical when learning from small datasets or partially observed, complex systems. Unfortunately, as the hybrid models become more flexible, the causal grounding provided by the mechanistic model can quickly be lost. We address this problem by leveraging another common source of domain knowledge: *ranking* of treatment effects for a set of interventions, even if the precise treatment effect is unknown. We encode this information in a *causal loss* that we combine with the standard predictive loss to arrive at a *hybrid loss* that biases our learning towards causally valid hybrid models. We demonstrate our ability to achieve a win-win, state-of-the-art predictive performance *and* causal validity, in the challenging task of modeling glucose dynamics post-exercise in individuals with type 1 diabetes.

Optimization · Discrete and Combinatorial Optimization

Jianan Zhou, Zhiguang Cao, Yaoxin Wu, Wen Song, Yining Ma, Jie Zhang, Xu Chi

Learning to solve vehicle routing problems (VRPs) has garnered much attention. However, most neural solvers are only structured and trained independently on a specific problem, making them less generic and practical. In this paper, we aim to develop a unified neural solver that can cope with a range of VRP variants simultaneously. Specifically, we propose a multi-task vehicle routing solver with mixture-of-experts (MVMoE), which greatly enhances the model capacity without a proportional increase in computation. We further develop a hierarchical gating mechanism for the MVMoE, delivering a good trade-off between empirical performance and computational complexity. Experimentally, our method significantly promotes zero-shot generalization performance on 10 unseen VRP variants, and showcases decent results on the few-shot setting and real-world benchmark instances. We further conduct extensive studies on the effect of MoE configurations in solving VRPs, and observe the superiority of hierarchical gating when facing out-of-distribution data. The source code is available at: https://github.com/RoyalSkye/Routing-MVMoE.

Optimization · Discrete and Combinatorial Optimization

Vicente Balmaseda, Ying Xu, Yixin Cao, Nate Veldt

Cluster deletion is an NP-hard graph clustering objective with applications in computational biology and social network analysis, where the goal is to delete a minimum number of edges to partition a graph into cliques. We first provide a tighter analysis of two previous approximation algorithms, improving their approximation guarantees from 4 to 3. Moreover, we show that both algorithms can be derandomized in a surprisingly simple way, by greedily taking a vertex of maximum degree in an auxiliary graph and forming a cluster around it. One of these algorithms relies on solving a linear program. Our final contribution is to design a new and purely combinatorial approach for doing so that is far more scalable in theory and practice.

Tri Dao, Albert Gu

While Transformers have been the main architecture behind deep learning's success in language modeling, state-space models (SSMs) such as Mamba have recently been shown to match or outperform Transformers at small to medium scale. We show that these families of models are actually quite closely related, and develop a rich framework of theoretical connections between SSMs and variants of attention, connected through various decompositions of a well-studied class of structured *semiseparable matrices*. Our state space duality (SSD) framework allows us to design a new architecture (**Mamba-2**) whose core layer is an a refinement of Mamba's selective SSM that is 2-8$\times$ faster, while continuing to be competitive with Transformers on language modeling.

Probabilistic Methods · Variational Inference

Diana Cai, Chirag Modi, Loucas Pillaud-Vivien, Charles Margossian, Robert Gower, David Blei, Lawrence Saul

Most leading implementations of black-box variational inference (BBVI) are based on optimizing a stochastic evidence lower bound (ELBO). But such approaches to BBVI often converge slowly due to the high variance of their gradient estimates and their sensitivity to hyperparameters. In this work, we propose _batch and match_ (BaM), an alternative approach to BBVI based on a score-based divergence. Notably, this score-based divergence can be optimized by a closed-form proximal update for Gaussian variational families with full covariance matrices. We analyze the convergence of BaM when the target distribution is Gaussian, and we prove that in the limit of infinite batch size the variational parameter updates converge exponentially quickly to the target mean and covariance. We also evaluate the performance of BaM on Gaussian and non-Gaussian target distributions that arise from posterior inference in hierarchical and deep generative models. In these experiments, we find that BaM typically converges in fewer (and sometimes significantly fewer) gradient evaluations than leading implementations of BBVI based on ELBO maximization.

Deep Learning · Algorithms

Mohammad Pezeshki, Diane Bouchacourt, Mark Ibrahim, Nicolas Ballas, Pascal Vincent, David Lopez-Paz

Environment annotations are essential for the success of many out-of-distribution (OOD) generalization methods. Unfortunately, these are costly to obtain and often limited by human annotators' biases. To achieve robust generalization, it is essential to develop algorithms for automatic environment discovery within datasets. Current proposals, which divide examples based on their training error, suffer from one fundamental problem. These methods introduce hyper-parameters and early-stopping criteria, which require a validation set with human-annotated environments, the very information subject to discovery. In this paper, we propose Cross-Risk Minimization (XRM) to address this issue. XRM trains twin networks, each learning from one random half of the training data, while imitating confident held-out mistakes made by its sibling. XRM provides a recipe for hyper-parameter tuning, does not require early-stopping, and can discover environments for all training and validation data. Algorithms built on top of XRM environments achieve oracle worst-group-accuracy, addressing a long-standing challenge in OOD generalization.

Social Aspects · Trustworthy Machine Learning

Deng-Bao Wang, Min-Ling Zhang

Although deep neural networks have achieved remarkable success, they often exhibit a significant deficiency in reliable uncertainty calibration. This paper focus on model calibratability, which assesses how amenable a model is to be well recalibrated post-hoc. We find that the widely used weight decay regularizer detrimentally affects model calibratability, subsequently leading to a decline in final calibration performance after post-hoc calibration. To identify the underlying causes leading to poor calibratability, we delve into the calibratability of intermediate features across the hidden layers. We observe a U-shaped trend in the calibratability of intermediate features from the bottom to the top layers, which indicates that over-compression of the top representation layers significantly hinders model calibratability. Based on the observations, this paper introduces a weak classifier hypothesis, i.e., given a weak classification head that has not been over-trained, the representation module can be better learned to produce more calibratable features. Consequently, we propose a progressively layer-peeled training (PLP) method to exploit this hypothesis, thereby enhancing model calibratability. Our comparative experiments show the effectiveness of our method, which improves model calibration and also yields competitive predictive performance.