ICML 2024 论文列表

Federated learning (FL) is an emerging machine learning paradigm for preserving data privacy. However, diverse client hardware often has varying computation resources. Such system heterogeneity limits the participation of resource-constrained clients in FL, and hence degrades the global model accuracy. To enable heterogeneous clients to participate in and contribute to FL training, previous works tackle this problem by assigning customized sub-models to individual clients with model pruning, distillation, or low-rank based techniques. Unfortunately, the global model trained by these methods still encounters performance degradation due to heterogeneous sub-model aggregation. Besides, most methods are heuristic-based and lack convergence analysis. In this work, we propose the FedLMT framework to bridge the performance gap, by assigning clients with a homogeneous pre-factorized low-rank model to substantially reduce resource consumption without conducting heterogeneous aggregation. We theoretically prove that the convergence of the low-rank model can guarantee the convergence of the original full model. To further meet clients' personalized resource needs, we extend FedLMT to pFedLMT, by separating model parameters into common and custom ones. Finally, extensive experiments are conducted to verify our theoretical analysis and show that FedLMT and pFedLMT outperform other baselines with much less communication and computation costs.

Class-Incremental Learning (CIL) seeks to learn new concepts without forgetting previously learned knowledge. To achieve this, rehearsal-based methods keep a replay memory consisting of a small number of trained samples from previous tasks. However, recent studies show that rehearsal-based methods are prone to overfitting on rehearsal samples, resulting in poor generalization on previous tasks. Since the generalization error is bounded by the margin on the training dataset, in this paper, we study the generalization by all-layer margin on deep neural networks to alleviate catastrophic forgetting. Specifically, we show that the average margin of the rehearsal samples are smaller during incremental learning. To acquire larger margin thus better generalization on rehearsal samples, we propose Multi-layer Rehearsal Feature Augmentation (MRFA) in rehearsal training to optimize the all-layer margin on rehearsal samples. The proposed method augments the features of rehearsal samples at each layer by gradient ascent step of the current model with respect to the feature. With such augmentations on layer features, the margin on rehearsal samples are larger, rehearsal samples are able to provide more information for refining the decision boundary during incremental learning, thus alleviating catastrophic forgetting. Extensive experiments show the effectiveness of MRFA on various CIL scenarios.

The learning to defer (L2D) framework allows models to defer their decisions to human experts. For L2D, the Bayes optimality is the basic requirement of theoretical guarantees for the design of consistent surrogate loss functions, which requires the minimizer (i.e., learned classifier) by the surrogate loss to be the Bayes optimality. However, we find that the original form of Bayes optimality fails to consider the dependence between the model and the expert, and such a dependence could be further exploited to design a better consistent loss for L2D. In this paper, we provide a new formulation for the Bayes optimality called dependent Bayes optimality, which reveals the dependence pattern in determining whether to defer. Based on the dependent Bayes optimality, we further present a deferral principle for L2D. Following the guidance of the deferral principle, we propose a novel consistent surrogate loss. Comprehensive experimental results on both synthetic and real-world datasets demonstrate the superiority of our proposed method.

Large Language Models (LLMs), renowned for their remarkable performance across diverse domains, present a challenge due to their colossal model size when it comes to practical deployment. In response to this challenge, efforts have been directed toward the application of traditional network pruning techniques to LLMs, uncovering a massive number of parameters can be pruned in one-shot without hurting performance. Building upon insights gained from pre-LLM models, particularly BERT-level language models, prevailing LLM pruning strategies have consistently adhered to the practice of uniformly pruning all layers at equivalent sparsity levels, resulting in robust performance. However, this observation stands in contrast to the prevailing trends observed in the field of vision models, where non-uniform layerwise sparsity typically yields substantially improved results. To elucidate the underlying reasons for this disparity, we conduct a comprehensive analysis of the distribution of token features within LLMs. In doing so, we discover a strong correlation with the emergence of outliers, defined as features exhibiting significantly greater magnitudes compared to their counterparts in feature dimensions. Inspired by this finding, we introduce a novel LLM pruning methodology that incorporates a tailored set of **non-uniform layerwise sparsity ratios** specifically designed for LLM pruning, termed as **O**utlier **W**eighed **L**ayerwise sparsity (**OWL**). The sparsity ratio of OWL is directly proportional to the outlier ratio observed within each layer, facilitating a more effective alignment between layerwise weight sparsity and outlier ratios. Our empirical evaluation, conducted across the LLaMA-V1/V2, Vicuna, OPT, and Mistral, spanning various benchmarks, demonstrates the distinct advantages offered by OWL over previous methods. For instance, OWL exhibits a remarkable performance gain, surpassing the state-of-the-art Wanda and SparseGPT by **61.22** and **6.80** perplexity at a high sparsity level of 70%, respectively, while delivering **2.6$\times$** end-to-end inference speed-up in the DeepSparse inference engine. Code is available at https://github.com/luuyin/OWL.git.

Many neural network architectures are known to be Turing Complete, and can thus, in principle implement arbitrary algorithms. However, Transformers are unique in that they can implement gradient-based learning algorithms *under simple parameter configurations*. This paper provides theoretical and empirical evidence that (non-linear) Transformers naturally learn to implement gradient descent *in function space*, which in turn enable them to learn non-linear functions in context. Our results apply to a broad class of combinations of non-linear architectures and non-linear in-context learning tasks. Additionally, we show that the optimal choice of non-linear activation depends in a natural way on the class of functions that need to be learned.

In recent years, there has been considerable interest in developing machine learning models on graphs to account for topological inductive biases. In particular, recent attention has been given to Gaussian processes on such structures since they can additionally account for uncertainty. However, graphs are limited to modelling relations between two vertices. In this paper, we go beyond this dyadic setting and consider polyadic relations that include interactions between vertices, edges and one of their generalisations, known as cells. Specifically, we propose Gaussian processes on cellular complexes, a generalisation of graphs that captures interactions between these higher-order cells. One of our key contributions is the derivation of two novel kernels, one that generalises the graph Matérn kernel and one that additionally mixes information of different cell types.

Various acceleration approaches for Policy Gradient (PG) have been analyzed within the realm of Reinforcement Learning (RL). However, the theoretical understanding of the widely used momentum-based acceleration method on PG remains largely open. In response to this gap, we adapt the celebrated Nesterov's accelerated gradient (NAG) method to policy optimization in RL, termed *Accelerated Policy Gradient* (APG). To demonstrate the potential of APG in achieving fast convergence, we formally prove that with the true gradient and under the softmax policy parametrization, APG converges to an optimal policy at rates: (i) $\tilde{O}(1/t^2)$ with nearly constant step sizes; (ii) $O(e^{-ct})$ with time-varying step sizes. To the best of our knowledge, this is the first characterization of the convergence rates of NAG in the context of RL. Notably, our analysis relies on one interesting finding: Regardless of the parameter initialization, APG ends up entering a locally nearly-concave regime, where APG can significantly benefit from the momentum, within finite iterations. Through numerical validation and experiments on the Atari 2600 benchmarks, we confirm that APG exhibits a $\tilde{O}(1/t^2)$ rate with nearly constant step sizes and a linear convergence rate with time-varying step sizes, significantly improving convergence over the standard PG.

We present a theoretical explanation of the “grokking” phenomenon (Power et al., 2022), where a model generalizes long after overfitting, for the originally-studied problem of modular addition. First, we show that early in gradient descent, so that the “kernel regime” approximately holds, no permutation-equivariant model can achieve small population error on modular addition unless it sees at least a constant fraction of all possible data points. Eventually, however, models escape the kernel regime. We show that one-hidden-layer quadratic networks that achieve zero training loss with bounded $\ell_\infty$ norm generalize well with substantially fewer training points, and further show such networks exist and can be found by gradient descent with small $\ell_\infty$ regularization. We further provide empirical evidence that these networks leave the kernel regime only after initially overfitting. Taken together, our results strongly support the case for grokking as a consequence of the transition from kernel-like behavior to limiting behavior of gradient descent on deep networks.

When using an autoencoder to learn the low-dimensional manifold of high-dimensional data, it is crucial to find the latent representations that preserve the geometry of the data manifold. However, most existing studies assume a Euclidean nature for the high-dimensional data space, which is arbitrary and often does not precisely reflect the underlying semantic or domain-specific attributes of the data. In this paper, we propose a novel autoencoder regularization framework based on the premise that the geometry of the data manifold can often be better captured with a well-designed similarity graph associated with data points. Given such a graph, we utilize a Riemannian geometric distortion measure as a regularizer to preserve the geometry derived from the graph Laplacian and make it suitable for larger-scale autoencoder training. Through extensive experiments, we show that our method outperforms existing state-of-the-art geometry-preserving and graph-based autoencoders with respect to learning accurate latent structures that preserve the graph geometry, and is particularly effective in learning dynamics in the latent space. Code is available at https://github.com/JungbinLim/GGAE-public.

Learning new knowledge frequently occurs in our dynamically changing world, e.g., humans culturally evolve by continuously acquiring new abilities to sustain their survival, leveraging collective intelligence rather than a large number of individual attempts. The effective learning paradigm during cultural evolution is termed socialized learning (SL). Consequently, a straightforward question arises: Can multi-agent systems acquire more new abilities like humans? In contrast to most existing methods that address continual learning and multi-agent collaboration, our emphasis lies in a more challenging problem: we prioritize the knowledge in the original expert classes, and as we adeptly learn new ones, the accuracy in the original expert classes stays superior among all in a directional manner. Inspired by population genetics and cognitive science, leading to unique and complete development, we propose Multi-Agent Socialized Collaboration (MASC), which achieves SL through interactions among multiple agents. Specifically, we introduce collective collaboration and reciprocal altruism modules, organizing collaborative behaviors, promoting information sharing, and facilitating learning and knowledge interaction among individuals. We demonstrate the effectiveness of multi-agent collaboration in an extensive empirical study. Our code will be publicly available at https://github.com/yxjdarren/SL.

Lagrangian Relaxation stands among the most efficient approaches for solving Mixed Integer Linear Programs (MILPs) with difficult constraints. Given any duals for these constraints, called Lagrangian Multipliers (LMs), it returns a bound on the optimal value of the MILP, and Lagrangian methods seek the LMs giving the best such bound. But these methods generally rely on iterative algorithms resembling gradient descent to maximize the concave piecewise linear dual function: the computational burden grows quickly with the number of relaxed constraints. We introduce a deep learning approach that bypasses the descent, effectively amortizing per instance optimization. A probabilistic encoder based on a graph neural network computes, given a MILP instance and its Continuous Relaxation (CR) solution, high-dimensional representations of relaxed constraints, which are turned into LMs by a decoder. We train the encoder and the decoder jointly by directly optimizing the bound obtained from the predicted multipliers. Our method is applicable to any problem with a compact MILP formulation, and to any Lagrangian Relaxation providing a tighter bound than CR. Experiments on two widely known problems, Multi-Commodity Network Design and Generalized Assignment, show that our approach closes up to 85% of the gap between the continuous relaxation and the best Lagrangian bound, and provides a high-quality warm-start for descent-based Lagrangian methods.

Current language models decode text token by token according to probabilistic distribution, and determining the appropriate candidates for the next token is crucial to ensure generation quality. This study introduces adaptive decoding, a mechanism that dynamically empowers language models to ascertain a sensible candidate set during generation. Specifically, we introduce an entropy-based metric called confidence and conceptualize determining the optimal candidate set as a confidence-increasing process. The rationality of including a token in the candidate set is assessed by leveraging the increment of confidence. Experimental results reveal that our method balances diversity and coherence well. The human evaluation shows that our method can generate human-preferred text. Additionally, our method can potentially improve the reasoning ability of language models.

The computational efficiency of many neural operators, widely used for learning solutions of PDEs, relies on the fast Fourier transform (FFT) for performing spectral computations. As the FFT is limited to equispaced (rectangular) grids, this limits the efficiency of such neural operators when applied to problems where the input and output functions need to be processed on general non-equispaced point distributions. Leveraging the observation that a limited set of Fourier (Spectral) modes suffice to provide the required expressivity of a neural operator, we propose a simple method, based on the efficient direct evaluation of the underlying spectral transformation, to extend neural operators to arbitrary domains. An efficient implementation of such *direct spectral evaluations* is coupled with existing neural operator models to allow the processing of data on arbitrary non-equispaced distributions of points. With extensive empirical evaluation, we demonstrate that the proposed method allows us to extend neural operators to arbitrary point distributions with significant gains in training speed over baselines, while retaining or improving the accuracy of Fourier neural operators (FNOs) and related neural operators.

The self-attention mechanism prevails in modern machine learning. It has an interesting functionality of adaptively selecting tokens from an input sequence by modulating the degree of attention localization, which many researchers speculate is the basis of the powerful model performance but complicates the underlying mechanism of the learning dynamics. In recent years, mainly two arguments have connected attention localization to the model performances. One is the rank collapse, where the embedded tokens by a self-attention block become very similar across different tokens, leading to a less expressive network. The other is the entropy collapse, where the attention probability approaches non-uniform and entails low entropy, making the learning dynamics more likely to be trapped in plateaus. These two failure modes may apparently contradict each other because the rank and entropy collapses are relevant to uniform and non-uniform attention, respectively. To this end, we characterize the notion of attention localization by the eigenspectrum of query-key parameter matrices and reveal that a small eigenspectrum variance leads attention to be localized. Interestingly, the small eigenspectrum variance prevents both rank and entropy collapse, leading to better model expressivity and trainability.

Despite the success of self-supervised pre-training in texts and images, applying it to multivariate time series (MTS) falls behind tailored methods for tasks like forecasting, imputation and anomaly detection. We propose a general-purpose framework, named UP2ME (**U**nivariate **P**re-training to **M**ultivariate Fin**e**-tuning). It conducts task-agnostic pre-training when downstream tasks are unspecified. Once the task and setting (e.g. forecasting length) are determined, it gives sensible solutions with frozen pre-trained parameters, which has not been achieved before. UP2ME is further refined by fine-tuning. A univariate-to-multivariate paradigm is devised to address the heterogeneity of temporal and cross-channel dependencies. In univariate pre-training, univariate instances with diverse lengths are generated for Masked AutoEncoder (MAE) pre-training, discarding cross-channel dependency. The pre-trained model handles downstream tasks by formulating them into specific mask-reconstruction problems. In multivariate fine-tuning, it constructs a dependency graph among channels using the pre-trained encoder to enhance cross-channel dependency capture. Experiments on eight real-world datasets show its SOTA performance in forecasting and imputation, approaching task-specific performance in anomaly detection. Our code is available at https://github.com/Thinklab-SJTU/UP2ME.

Transformers have become foundational architectures for both natural language and computer vision tasks. However, the high computational cost makes it quite challenging to deploy on resource-constraint devices. This paper investigates the computational bottleneck modules of efficient transformer, *i.e.*, normalization layers and attention modules. LayerNorm is commonly used in transformer architectures but is not computational friendly due to statistic calculation during inference. However, replacing LayerNorm with more efficient BatchNorm in transformer often leads to inferior performance and collapse in training. To address this problem, we propose a novel method named PRepBN to progressively replace LayerNorm with re-parameterized BatchNorm in training. Moreover, we propose a simplified linear attention (SLA) module that is simple yet effective to achieve strong performance. Extensive experiments on image classification as well as object detection demonstrate the effectiveness of our proposed method. For example, our SLAB-Swin obtains $83.6\%$ top-1 accuracy on ImageNet-1K with $16.2$ms latency, which is $2.4$ms less than that of Flatten-Swin with $0.1\%$ higher accuracy. We also evaluated our method for language modeling task and obtain comparable performance and lower latency. Codes are publicly available at https://github.com/xinghaochen/SLAB and https://github.com/mindspore-lab/models/tree/master/research/huawei-noah/SLAB.

Given *any* algorithm for convex optimization that uses exact first-order information (i.e., function values and subgradients), we show how to use such an algorithm to solve the problem with access to *inexact* first-order information. This is done in a ``black-box'' manner without knowledge of the internal workings of the algorithm. This complements previous work that considers the performance of specific algorithms like (accelerated) gradient descent with inexact information. In particular, our results apply to a wider range of algorithms beyond variants of gradient descent, e.g., projection-free methods, cutting-plane methods, or any other first-order methods formulated in the future. Further, they also apply to algorithms that handle structured nonconvexities like mixed-integer decision variables.

Recent advances in foundation models have led to a promising trend of developing large recommendation models to leverage vast amounts of available data. Still, mainstream models remain embarrassingly small in size and naive enlarging does not lead to sufficient performance gain, suggesting a deficiency in the model scalability. In this paper, we identify the embedding collapse phenomenon as the inhibition of scalability, wherein the embedding matrix tends to occupy a low-dimensional subspace. Through empirical and theoretical analysis, we demonstrate a two-sided effect of feature interaction specific to recommendation models. On the one hand, interacting with collapsed embeddings restricts embedding learning and exacerbates the collapse issue. On the other hand, interaction is crucial in mitigating the fitting of spurious features as a scalability guarantee. Based on our analysis, we propose a simple yet effective multi-embedding design incorporating embedding-set-specific interaction modules to learn embedding sets with large diversity and thus reduce collapse. Extensive experiments demonstrate that this proposed design provides consistent scalability and effective collapse mitigation for various recommendation models. Code is available at this repository: https://github.com/thuml/Multi-Embedding.

Self-attention and masked self-attention are at the heart of Transformers' outstanding success. Still, our mathematical understanding of attention, in particular of its Lipschitz properties — which are key when it comes to analyzing robustness and expressive power — is incomplete. We provide a detailed study of the Lipschitz constant of self-attention in several practical scenarios, discussing the impact of the sequence length $n$ and layer normalization on the local Lipschitz constant of both unmasked and masked self-attention. In particular, we show that for inputs of length $n$ in any compact set, the Lipschitz constant of self-attention is bounded by $\sqrt{n}$ up to a constant factor and that this bound is tight for reasonable sequence lengths. When the sequence length $n$ is too large for the previous bound to be tight, which we refer to as the mean-field regime, we provide an upper bound and a matching lower bound which are independent of $n$. Our mean-field framework for masked self-attention is novel and of independent interest. Our experiments on pretrained and randomly initialized BERT and GPT-2 support our theoretical findings.