ICML 2024 论文列表

We obtain the first positive results for bounded sample compression in the agnostic regression setting with the $\ell_p$ loss, where $p\in [1,\infty]$. We construct a generic approximate sample compression scheme for real-valued function classes exhibiting exponential size in the fat-shattering dimension but independent of the sample size. Notably, for linear regression, an approximate compression of size linear in the dimension is constructed. Moreover, for $\ell_1$ and $\ell_\infty$ losses, we can even exhibit an efficient exact sample compression scheme of size linear in the dimension. We further show that for every other $\ell_p$ loss, $p\in (1,\infty)$, there does not exist an exact agnostic compression scheme of bounded size. This refines and generalizes a negative result of David, Moran, and Yehudayoff (2016) for the $\ell_2$ loss. We close by posing general open questions: for agnostic regression with $\ell_1$ loss, does every function class admit an exact compression scheme of polynomial size in the pseudo-dimension? For the $\ell_2$ loss, does every function class admit an approximate compression scheme of polynomial size in the fat-shattering dimension? These questions generalize Warmuth's classic sample compression conjecture for realizable-case classification (Warmuth, 2003).

This work presents orthogonal attention for constructing neural operators to serve as surrogates to model the solutions of a family of Partial Differential Equations (PDEs). The motivation is that the kernel integral operator, which is usually at the core of neural operators, can be reformulated with orthonormal eigenfunctions. Inspired by the success of the neural approximation of eigenfunctions (Deng et al., 2022), we opt to directly parameterize the involved eigenfunctions with flexible neural networks (NNs), based on which the input function is then transformed by the rule of kernel integral. Surprisingly, the resulting NN module bears a striking resemblance to regular attention mechanisms, albeit without softmax. Instead, it incorporates an orthogonalization operation that provides regularization during model training and helps mitigate overfitting, particularly in scenarios with limited data availability. In practice, the orthogonalization operation can be implemented with minimal additional overheads. Experiments on six standard neural operator benchmark datasets comprising both regular and irregular geometries show that our method can outperform competing baselines with decent margins.

In federated learning (FL), the multi-step update and data heterogeneity among clients often lead to a loss landscape with sharper minima, degenerating the performance of the resulted global model. Prevalent federated approaches incorporate sharpness-aware minimization (SAM) into local training to mitigate this problem. However, the local loss landscapes may not accurately reflect the flatness of global loss landscape in heterogeneous environments; as a result, minimizing local sharpness and calculating perturbations on client data might not align the efficacy of SAM in FL with centralized training. To overcome this challenge, we propose FedLESAM, a novel algorithm that locally estimates the direction of global perturbation on client side as the difference between global models received in the previous active and current rounds. Besides the improved quality, FedLESAM also speed up federated SAM-based approaches since it only performs once backpropagation in each iteration. Theoretically, we prove a slightly tighter bound than its original FedSAM by ensuring consistent perturbation. Empirically, we conduct comprehensive experiments on four federated benchmark datasets under three partition strategies to demonstrate the superior performance and efficiency of FedLESAM.

Pre-trained Vision-Language Models (VLMs) are able to understand visual concepts, describe and decompose complex tasks into sub-tasks, and provide feedback on task completion. In this paper, we aim to leverage these capabilities to support the training of reinforcement learning (RL) agents. In principle, VLMs are well suited for this purpose, as they can naturally analyze image-based observations and provide feedback (reward) on learning progress. However, inference in VLMs is computationally expensive, so querying them frequently to compute rewards would significantly slowdown the training of an RL agent. To address this challenge, we propose a framework named Code as Reward (VLM-CaR). VLM-CaR produces dense reward functions from VLMs through code generation, thereby significantly reducing the computational burden of querying the VLM directly. We show that the dense rewards generated through our approach are very accurate across a diverse set of discrete and continuous environments, and can be more effective in training RL policies than the original sparse environment rewards.

We study the problem of parameter-free stochastic optimization, inquiring whether, and under what conditions, do fully parameter-free methods exist: these are methods that achieve convergence rates competitive with optimally tuned methods, without requiring significant knowledge of the true problem parameters. Existing parameter-free methods can only be considered ``partially'' parameter-free, as they require some non-trivial knowledge of the true problem parameters, such as a bound on the stochastic gradient norms, a bound on the distance to a minimizer, etc. In the non-convex setting, we demonstrate that a simple hyperparameter search technique results in a fully parameter-free method that outperforms more sophisticated state-of-the-art algorithms. We also provide a similar result in the convex setting with access to noisy function values under mild noise assumptions. Finally, assuming only access to stochastic gradients, we establish a lower bound that renders fully parameter-free stochastic convex optimization infeasible, and provide a method which is (partially) parameter-free up to the limit indicated by our lower bound.

This paper studies the causal relations from subsampled time series, in which measurements are sparse and sampled at a coarser timescale than the causal timescale of the underlying system. In such data, because there are numerous missing time-slices (i.e., cross-sections at each time point) between two consecutive measurements, conventional causal discovery methods designed for standard time series data would produce significant errors. To learn causal relations from subsampled time series, a typical solution is to conduct different interventions and then make a comparison. However, full interventions are often expensive, unethical, or even infeasible, particularly in fields such as health and social science. In this paper, we first explore how readily available two-time-slices data can replace intervention data to improve causal ordering, and propose a novel Descendant Hierarchical Topology algorithm with Conditional Independence Test (DHT-CIT) to learn causal relations from subsampled time series using only two time-slices. Specifically, we develop a conditional independence criterion that can be applied iteratively to test each node from time series and identify all of its descendant nodes. Empirical results on both synthetic and real-world datasets demonstrate the superiority of our DHT-CIT algorithm.

Large language models (LLMs) can store a vast amount of world knowledge, often extractable via question-answering (e.g., "What is Abraham Lincoln's birthday?''). However, do they answer such questions based on exposure to similar questions during training (i.e., cheating), or by genuinely learning to extract knowledge from sources like Wikipedia? In this paper, we investigate this issue using a controlled biography dataset. We find a strong correlation between the model's ability to extract knowledge and various _diversity measures_ of the training data. **Essentially**, for knowledge to be reliably extracted, it must be sufficiently augmented (e.g., through paraphrasing, sentence shuffling) _during pretraining_. Without such augmentation, knowledge may be memorized but not extractable, leading to 0% accuracy, regardless of subsequent instruction fine-tuning. To understand why this occurs, we employ (nearly) linear probing to demonstrate a strong connection between the observed correlation and _how the model internally encodes knowledge_ --- whether it is linearly encoded in the hidden embeddings of entity names or distributed across other token embeddings in the training text. **This paper provides several key recommendations for LLM pretraining in the industry:** (1) rewrite the pretraining data --- using small, auxiliary models --- to provide knowledge augmentation, and (2) incorporate more instruction-finetuning data into the pretraining stage before it becomes too late.

To enable closed form conditioning, a common assumption in Gaussian process (GP) regression is independent and identically distributed Gaussian observation noise. This strong and simplistic assumption is often violated in practice, which leads to unreliable inferences and uncertainty quantification. Unfortunately, existing methods for robustifying GPs break closed-form conditioning, which makes them less attractive to practitioners and significantly more computationally expensive. In this paper, we demonstrate how to perform provably robust and conjugate Gaussian process (RCGP) regression at virtually no additional cost using generalised Bayesian inference. RCGP is particularly versatile as it enables exact conjugate closed form updates in all settings where standard GPs admit them. To demonstrate its strong empirical performance, we deploy RCGP for problems ranging from Bayesian optimisation to sparse variational Gaussian processes.

Learning to defer with multiple experts is a framework where the learner can choose to defer the prediction to several experts. While this problem has received significant attention in classification contexts, it presents unique challenges in regression due to the infinite and continuous nature of the label space. In this work, we introduce a novel framework of *regression with deferral*, which involves deferring the prediction to multiple experts. We present a comprehensive analysis for both the single-stage scenario, where there is simultaneous learning of predictor and deferral functions, and the two-stage scenario, which involves a pre-trained predictor with a learned deferral function. We introduce new surrogate loss functions for both scenarios and prove that they are supported by $H$-consistency bounds. These bounds provide consistency guarantees that are stronger than Bayes consistency, as they are non-asymptotic and hypothesis set-specific. Our framework is versatile, applying to multiple experts, accommodating any bounded regression losses, addressing both instance-dependent and label-dependent costs, and supporting both single-stage and two-stage methods. Our single-stage formulation subsumes as a special case the recent *regression with abstention* (Cheng et al., 2023) framework, where only a single expert is considered, specifically for the squared loss and a label-independent cost. Minimizing our proposed loss functions directly leads to novel algorithms for regression with deferral. We report the results of extensive experiments showing the effectiveness of our proposed algorithms.

Fine-tuning is a widespread technique that allows practitioners to transfer pre-trained capabilities, as recently showcased by the successful applications of foundation models. However, fine-tuning reinforcement learning (RL) models remains a challenge. This work conceptualizes one specific cause of poor transfer, accentuated in the RL setting by the interplay between actions and observations: *forgetting of pre-trained capabilities*. Namely, a model deteriorates on the state subspace of the downstream task not visited in the initial phase of fine-tuning, on which the model behaved well due to pre-training. This way, we lose the anticipated transfer benefits. We identify conditions when this problem occurs, showing that it is common and, in many cases, catastrophic. Through a detailed empirical analysis of the challenging NetHack and Montezuma's Revenge environments, we show that standard knowledge retention techniques mitigate the problem and thus allow us to take full advantage of the pre-trained capabilities. In particular, in NetHack, we achieve a new state-of-the-art for neural models, improving the previous best score from $5$K to over $10$K points in the Human Monk scenario.

We study an online contextual dynamic pricing problem, where customers decide whether to purchase a product based on its features and price. We introduce a novel approach to modeling a customer's expected demand by incorporating feature-based price elasticity, which can be equivalently represented as a valuation with heteroscedastic noise. To solve the problem, we propose a computationally efficient algorithm called "Pricing with Perturbation (PwP)", which enjoys an $O(\sqrt{dT\log T})$ regret while allowing arbitrary adversarial input context sequences. We also prove a matching lower bound at $\Omega(\sqrt{dT})$ to show the optimality regarding $d$ and $T$ (up to $\log T$ factors). Our results shed light on the relationship between contextual elasticity and heteroscedastic valuation, providing insights for effective and practical pricing strategies.

*Visual reprogramming* (VR) is a prompting technique that aims to re-purpose a pre-trained model (e.g., a classifier on ImageNet) to target tasks (e.g., medical data prediction) by learning a *small-scale pattern* added into input images instead of tuning considerable parameters within the model. The location of the pattern within input samples is usually determined by a pre-defined mask *shared across all samples*. In this paper, we show that the shared mask potentially limits VR's generalization and increases its approximation error due to the lack of sample-level adaptation. Motivated by this finding, we design a new framework for VR called *sample-specific multi-channel masks* (SMM). Specifically, SMM employs a lightweight ConvNet and patch-wise interpolation to generate sample-specific three-channel masks instead of a shared and pre-defined mask. Since we generate different masks for individual samples, SMM is theoretically shown to reduce approximation error for the target tasks compared with existing state-of-the-art VR methods. We also empirically demonstrate its performance gain on both ResNet and ViT. The success of SMM further highlights the broader applicability of VR in leveraging the latent knowledge of pre-trained models for various target tasks. Our code is available at https://github.com/tmlr-group/SMM.

This paper explores the size-invariance of evaluation metrics in Salient Object Detection (SOD), especially when multiple targets of diverse sizes co-exist in the same image. We observe that current metrics are size-sensitive, where larger objects are focused, and smaller ones tend to be ignored. We argue that the evaluation should be size-invariant because bias based on size is unjustified without additional semantic information. In pursuit of this, we propose a generic approach that evaluates each salient object separately and then combines the results, effectively alleviating the imbalance. We further develop an optimization framework tailored to this goal, achieving considerable improvements in detecting objects of different sizes. Theoretically, we provide evidence supporting the validity of our new metrics and present the generalization analysis of SOD. Extensive experiments demonstrate the effectiveness of our method.

We revisit the objective perturbations framework for differential privacy where noise is added to the input $A\in \mathcal{S}$ and the result is then projected back to the space of admissible datasets $\mathcal{S}$. Through this framework, we first design novel efficient algorithms to privately release pair-wise cosine similarities. Second, we derive a novel algorithm to compute $k$-way marginal queries over $n$ features. Prior work could achieve comparable guarantees only for $k$ even. Furthermore, we extend our results to $t$-sparse datasets, where our efficient algorithms yields novel, stronger guarantees whenever $t\le n^{5/6}/\log n.$ Finally, we provide a theoretical perspective on why *fast* input perturbation algorithms works well in practice. The key technical ingredients behind our results are tight sum-of-squares certificates upper bounding the Gaussian complexity of sets of solutions.

New capabilities in foundation models are owed in large part to massive, widely-sourced, and under-documented training data collections. Existing practices in data collection have led to challenges in tracing authenticity, verifying consent, preserving privacy, addressing representation and bias, respecting copyright, and overall developing ethical and trustworthy foundation models. In response, regulation is emphasizing the need for training data transparency to understand foundation models’ limitations. Based on a large-scale analysis of the foundation model training data landscape and existing solutions, we identify the missing infrastructure to facilitate responsible foundation model development practices. We examine the current shortcomings of common tools for tracing data authenticity, consent, and documentation, and outline how policymakers, developers, and data creators can facilitate responsible foundation model development by adopting universal data provenance standards.

A main theoretical puzzle is why over-parameterized Neural Networks (NNs) generalize well when trained to zero loss (i.e., so they interpolate the data). Usually, the NN is trained with Stochastic Gradient Descent (SGD) or one of its variants. However, recent empirical work examined the generalization of a random NN that interpolates the data: the NN was sampled from a seemingly uniform prior over the parameters, conditioned on that the NN perfectly classifying the training set. Interestingly, such a NN sample typically generalized as well as SGD-trained NNs. We prove that such a random NN interpolator typically generalizes well if there exists an underlying narrow ``teacher NN" that agrees with the labels. Specifically, we show that such a `flat' prior over the NN parametrization induces a rich prior over the NN functions, due to the redundancy in the NN structure. In particular, this creates a bias towards simpler functions, which require less relevant parameters to represent --- enabling learning with a sample complexity approximately proportional to the complexity of the teacher (roughly, the number of non-redundant parameters), rather than the student's.

We study the classic problem of correlation clustering in dynamic vertex streams. In this setting, vertices are either added or randomly deleted over time, and each vertex pair is connected by a positive or negative edge. The objective is to continuously find a partition which minimizes the sum of positive edges crossing clusters and negative edges within clusters. We present an algorithm that maintains an $O(1)$-approximation with $O(\text{polylog} n)$ amortized update time. Prior to our work Behnezhad et al. in SODA 2023 achieved a $5$-approximation with $O(1)$ expected update time in edge streams which translates in vertex streams to an $O(D)$-update time where $D$ is the maximum possible degree. Finally we complement our theoretical analysis with experiments on real world data.

Bayesian optimization (BO) is a popular method to optimize costly black-box functions, and meta-learning has emerged as a way to leverage knowledge from related tasks to optimize new tasks faster. However, existing meta-learning methods for BO rely on surrogate models that are not scalable or are sensitive to varying input scales and noise types across tasks. Moreover, they often overlook the uncertainty associated with task similarity, leading to unreliable task adaptation when a new task differs significantly or has not been sufficiently explored yet. We propose a novel meta-learning BO approach that bypasses the surrogate model and directly learns the utility of queries across tasks. It explicitly models task uncertainty and includes an auxiliary model to enable robust adaptation to new tasks. Extensive experiments show that our method achieves strong performance and outperforms multiple meta-learning BO methods across various benchmarks.

In recent years, the state-of-the-art in deep learning has been dominated by very large models that have been pre-trained on vast amounts of data. The paradigm is very simple: investing more computational resources (optimally) leads to better performance, and even predictably so; neural scaling laws have been derived that accurately forecast the performance of a network for a desired level of compute. This leads to the notion of a 'compute-optimal' model, i.e. a model that allocates a given level of compute during training optimally to maximize performance. In this work, we extend the concept of optimality by allowing for an 'adaptive' model, i.e. a model that can change its shape during training. By doing so, we can design adaptive models that optimally traverse between the underlying scaling laws and outpace their `static' counterparts, leading to a significant reduction in the required compute to reach a given target performance. We show that our approach generalizes across modalities and different shape parameters.