PyTorchKR Papers

최신 논문

총 638편
태그·기간 등 상세 필터

카테고리 탐색

로그인 후 이 태그 구독

arXiv:2410.08977

Online-to-PAC generalization bounds under graph-mixing dependencies

Traditional generalization results in statistical learning require a training data set made of independently drawn examples. Most of the recent efforts to relax this independence assumption have considered either purely temporal (mixing) dependencies, or graph-dependencies, where non-adjacent vertices correspond to independent random variables. Both approaches have their own limitations, the former requiring a temporal ordered structure, and the latter lacking a way to quantify the strength of inter-dependencies. In this work, we bridge these two lines of work by proposing a framework where dependencies decay with graph distance. We derive generalization bounds leveraging the online-to-PAC framework, by deriving a concentration result and introducing an online learning framework incorporating the graph structure. The resulting high-probability generalization guarantees depend on both the mixing rate and the graph's chromatic number.

arXiv:2410.08497

Towards Sharper Risk Bounds for Minimax Problems

Minimax problems have achieved success in machine learning such as adversarial training, robust optimization, reinforcement learning. For theoretical analysis, current optimal excess risk bounds, which are composed by generalization error and optimization error, present 1/n-rates in strongly-convex-strongly-concave (SC-SC) settings. Existing studies mainly focus on minimax problems with specific algorithms for optimization error, with only a few studies on generalization performance, which limit better excess risk bounds. In this paper, we study the generalization bounds measured by the gradients of primal functions using uniform localized convergence. We obtain a sharper high probability generalization error bound for nonconvex-strongly-concave (NC-SC) stochastic minimax problems. Furthermore, we provide dimension-independent results under Polyak-Lojasiewicz condition for the outer layer. Based on our generalization error bound, we analyze some popular algorithms such as empirical saddle point (ESP), gradient descent ascent (GDA) and stochastic gradient descent ascent (SGDA). We derive better excess primal risk bounds with further reasonable assumptions, which, to the best of our knowledge, are n times faster than exist results in minimax problems.

arXiv:2410.08026

Generalization Bounds and Model Complexity for Kolmogorov-Arnold Networks

Kolmogorov-Arnold Network (KAN) is a network structure recently proposed by Liu et al. (2024) that offers improved interpretability and a more parsimonious design in many science-oriented tasks compared to multi-layer perceptrons. This work provides a rigorous theoretical analysis of KAN by establishing generalization bounds for KAN equipped with activation functions that are either represented by linear combinations of basis functions or lying in a low-rank Reproducing Kernel Hilbert Space (RKHS). In the first case, the generalization bound accommodates various choices of basis functions in forming the activation functions in each layer of KAN and is adapted to different operator norms at each layer. For a particular choice of operator norms, the bound scales with the $l_1$ norm of the coefficient matrices and the Lipschitz constants for the activation functions, and it has no dependence on combinatorial parameters (e.g., number of nodes) outside of logarithmic factors. Moreover, our result does not require the boundedness assumption on the loss function and, hence, is applicable to a general class of regression-type loss functions. In the low-rank case, the generalization bound scales polynomially with the underlying ranks as well as the Lipschitz constants of the activation functions in each layer. These bounds are empirically investigated for KANs trained with stochastic gradient descent on simulated and real data sets. The numerical results demonstrate the practical relevance of these bounds.

arXiv:2410.07719

Understanding Adversarially Robust Generalization via Weight-Curvature Index

Despite extensive research on adversarial examples, the underlying mechanisms of adversarially robust generalization, a critical yet challenging task for deep learning, remain largely unknown. In this work, we propose a novel perspective to decipher adversarially robust generalization through the lens of the Weight-Curvature Index (WCI). The proposed WCI quantifies the vulnerability of models to adversarial perturbations using the Frobenius norm of weight matrices and the trace of Hessian matrices. We prove generalization bounds based on PAC-Bayesian theory and second-order loss function approximations to elucidate the interplay between robust generalization gap, model parameters, and loss landscape curvature. Our theory and experiments show that WCI effectively captures the robust generalization performance of adversarially trained models. By offering a nuanced understanding of adversarial robustness based on the scale of model parameters and the curvature of the loss landscape, our work provides crucial insights for designing more resilient deep learning models, enhancing their reliability and security.

arXiv:2410.01969

Which Algorithms Have Tight Generalization Bounds?

We study which machine learning algorithms have tight generalization bounds. First, we present conditions that preclude the existence of tight generalization bounds. Specifically, we show that algorithms that have certain inductive biases that cause them to be unstable do not admit tight generalization bounds. Next, we show that algorithms that are sufficiently stable do have tight generalization bounds. We conclude with a simple characterization that relates the existence of tight generalization bounds to the conditional variance of the algorithm's loss.

arXiv:2410.02247공식 코드 · Papers with Code

Theoretical Insights into Fine-Tuning Attention Mechanism: Generalization and Optimization

Large Language Models (LLMs), built on Transformer architectures, exhibit remarkable generalization across a wide range of tasks. However, fine-tuning these models for specific tasks remains resource-intensive due to their extensive parameterization. In this paper, we explore two remarkable phenomena related to the attention mechanism during the fine-tuning of LLMs (where $\mathbf{W}_q$, $\mathbf{W}_k$, and $\mathbf{W}_v$ denote the weights of the query, key, and value layers, respectively). The first phenomenon, termed "Unequal Importance of Attention Matrices", highlights the impact of fine-tuning different weight matrices. It shows that optimizing the $\mathbf{W}_v$ matrix yields significantly better performance than optimizing the $\mathbf{W}_k$ matrix. Fine-tuning only the $\mathbf{W}_q$ and $\mathbf{W}_v$ matrices is computationally efficient while delivering results comparable to, or even better than fine-tuning all three matrices ($\mathbf{W}_q$, $\mathbf{W}_k$, and $\mathbf{W}_v$). The second phenomenon,"Attention Matrices with Customized Learning Rate Lead to Better Convergence", emphasizes the importance of assigning distinct learning rates to these matrices. Specifically, a higher learning rate for the $\mathbf{W}_v$ matrix compared to $\mathbf{W}_q$ and $\mathbf{W}_k$ accelerates convergence and improves performance. Building on these insights, we propose a new strategy that improves fine-tuning efficiency in terms of both storage and time. Experimental results on benchmark datasets validate the effectiveness of this approach, supporting our theoretical findings. Our analysis lays the theoretical groundwork for configuring and improving algorithms in LLMs fine-tuning.

arXiv:2410.02724

Large Language Models as Markov Chains

Large language models (LLMs) are remarkably efficient across a wide range of natural language processing tasks and well beyond them. However, a comprehensive theoretical analysis of the LLMs' generalization capabilities remains elusive. In our paper, we approach this task by drawing an equivalence between autoregressive transformer-based language models and Markov chains defined on a finite state space. This allows us to study the multi-step inference mechanism of LLMs from first principles. We relate the obtained results to the pathological behavior observed with LLMs such as repetitions and incoherent replies with high temperature. Finally, we leverage the proposed formalization to derive pre-training and in-context learning generalization bounds for LLMs under realistic data and model assumptions. Experiments with the most recent Llama and Gemma herds of models show that our theory correctly captures their behavior in practice.

arXiv:2409.17932공식 코드 · Papers with Code

Sample Compression Unleashed: New Generalization Bounds for Real Valued Losses

The sample compression theory provides generalization guarantees for predictors that can be fully defined using a subset of the training dataset and a (short) message string, generally defined as a binary sequence. Previous works provided generalization bounds for the zero-one loss, which is restrictive notably when applied to deep learning approaches. In this paper, we present a general framework for deriving new sample compression bounds that hold for real-valued unbounded losses. Using the Pick-To-Learn (P2L) meta-algorithm, which transforms the training method of any machine-learning predictor to yield sample-compressed predictors, we empirically demonstrate the tightness of the bounds and their versatility by evaluating them on random forests and multiple types of neural networks.

arXiv:2409.00575

Online Optimization for Learning to Communicate over Time-Correlated Channels

Machine learning techniques have garnered great interest in designing communication systems owing to their capacity in tackling with channel uncertainty. To provide theoretical guarantees for learning-based communication systems, some recent works analyze generalization bounds for devised methods based on the assumption of Independently and Identically Distributed (I.I.D.) channels, a condition rarely met in practical scenarios. In this paper, we drop the I.I.D. channel assumption and study an online optimization problem of learning to communicate over time-correlated channels. To address this issue, we further focus on two specific tasks: optimizing channel decoders for time-correlated fading channels and selecting optimal codebooks for time-correlated additive noise channels. For utilizing temporal dependence of considered channels to better learn communication systems, we develop two online optimization algorithms based on the optimistic online mirror descent framework. Furthermore, we provide theoretical guarantees for proposed algorithms via deriving sub-linear regret bound on the expected error probability of learned systems. Extensive simulation experiments have been conducted to validate that our presented approaches can leverage the channel correlation to achieve a lower average symbol error rate compared to baseline methods, consistent with our theoretical findings.

arXiv:2409.05188

Learning to Classify Quantum Phases of Matter with a Few Measurements

We study the identification of quantum phases of matter, at zero temperature, when only part of the phase diagram is known in advance. Following a supervised learning approach, we show how to use our previous knowledge to construct an observable capable of classifying the phase even in the unknown region. By using a combination of classical and quantum techniques, such as tensor networks, kernel methods, generalization bounds, quantum algorithms, and shadow estimators, we show that, in some cases, the certification of new ground states can be obtained with a polynomial number of measurements. An important application of our findings is the classification of the phases of matter obtained in quantum simulators, e.g., cold atom experiments, capable of efficiently preparing ground states of complex many-particle systems and applying simple measurements, e.g., single qubit measurements, but unable to perform a universal set of gates.

arXiv:2409.02668

Introduction to Machine Learning

This book introduces the mathematical foundations and techniques that lead to the development and analysis of many of the algorithms that are used in machine learning. It starts with an introductory chapter that describes notation used throughout the book and serve at a reminder of basic concepts in calculus, linear algebra and probability and also introduces some measure theoretic terminology, which can be used as a reading guide for the sections that use these tools. The introductory chapters also provide background material on matrix analysis and optimization. The latter chapter provides theoretical support to many algorithms that are used in the book, including stochastic gradient descent, proximal methods, etc. After discussing basic concepts for statistical prediction, the book includes an introduction to reproducing kernel theory and Hilbert space techniques, which are used in many places, before addressing the description of various algorithms for supervised statistical learning, including linear methods, support vector machines, decision trees, boosting, or neural networks. The subject then switches to generative methods, starting with a chapter that presents sampling methods and an introduction to the theory of Markov chains. The following chapter describe the theory of graphical models, an introduction to variational methods for models with latent variables, and to deep-learning based generative models. The next chapters focus on unsupervised learning methods, for clustering, factor analysis and manifold learning. The final chapter of the book is theory-oriented and discusses concentration inequalities and generalization bounds.

arXiv:2408.11629

A Markovian Model for Learning-to-Optimize

We present a probabilistic model for stochastic iterative algorithms with the use case of optimization algorithms in mind. Based on this model, we present PAC-Bayesian generalization bounds for functions that are defined on the trajectory of the learned algorithm, for example, the expected (non-asymptotic) convergence rate and the expected time to reach the stopping criterion. Thus, not only does this model allow for learning stochastic algorithms based on their empirical performance, it also yields results about their actual convergence rate and their actual convergence time. We stress that, since the model is valid in a more general setting than learning-to-optimize, it is of interest for other fields of application, too. Finally, we conduct five practically relevant experiments, showing the validity of our claims.

arXiv:2408.08675

Misclassification excess risk bounds for PAC-Bayesian classification via convexified loss

PAC-Bayesian bounds have proven to be a valuable tool for deriving generalization bounds and for designing new learning algorithms in machine learning. However, it typically focus on providing generalization bounds with respect to a chosen loss function. In classification tasks, due to the non-convex nature of the 0-1 loss, a convex surrogate loss is often used, and thus current PAC-Bayesian bounds are primarily specified for this convex surrogate. This work shifts its focus to providing misclassification excess risk bounds for PAC-Bayesian classification when using a convex surrogate loss. Our key ingredient here is to leverage PAC-Bayesian relative bounds in expectation rather than relying on PAC-Bayesian bounds in probability. We demonstrate our approach in several important applications.

arXiv:2408.07305

Learning Decisions Offline from Censored Observations with ε-insensitive Operational Costs

Many important managerial decisions are made based on censored observations. Making decisions without adequately handling the censoring leads to inferior outcomes. We investigate the data-driven decision-making problem with an offline dataset containing the feature data and the censored historical data of the variable of interest without the censoring indicators. Without assuming the underlying distribution, we design and leverage {\epsilon}-insensitive operational costs to deal with the unobserved censoring in an offline data-driven fashion. We demonstrate the customization of the {\epsilon}-insensitive operational costs for a newsvendor problem and use such costs to train two representative ML models, including linear regression (LR) models and neural networks (NNs). We derive tight generalization bounds for the custom LR model without regularization (LR-{\epsilon}NVC) and with regularization (LR-{\epsilon}NVC-R), and a high-probability generalization bound for the custom NN (NN-{\epsilon}NVC) trained by stochastic gradient descent. The theoretical results reveal the stability and learnability of LR-{\epsilon}NVC, LR-{\epsilon}NVC-R and NN-{\epsilon}NVC. We conduct extensive numerical experiments to compare LR-{\epsilon}NVC-R and NN-{\epsilon}NVC with two existing approaches, estimate-as-solution (EAS) and integrated estimation and optimization (IEO). The results show that LR-{\epsilon}NVC-R and NN-{\epsilon}NVC outperform both EAS and IEO, with maximum cost savings up to 14.40% and 12.21% compared to the lowest cost generated by the two existing approaches. In addition, LR-{\epsilon}NVC-R's and NN-{\epsilon}NVC's order quantities are statistically significantly closer to the optimal solutions should the underlying distribution be known.

arXiv:2407.19715

Generalization bounds for regression and classification on adaptive covering input domains

Our main focus is on the generalization bound, which serves as an upper limit for the generalization error. Our analysis delves into regression and classification tasks separately to ensure a thorough examination. We assume the target function is real-valued and Lipschitz continuous for regression tasks. We use the 2-norm and a root-mean-square-error (RMSE) variant to measure the disparities between predictions and actual values. In the case of classification tasks, we treat the target function as a one-hot classifier, representing a piece-wise constant function, and employ 0/1 loss for error measurement. Our analysis underscores the differing sample complexity required to achieve a concentration inequality of generalization bounds, highlighting the variation in learning efficiency for regression and classification tasks. Furthermore, we demonstrate that the generalization bounds for regression and classification functions are inversely proportional to a polynomial of the number of parameters in a network, with the degree depending on the hypothesis class and the network architecture. These findings emphasize the advantages of over-parameterized networks and elucidate the conditions for benign overfitting in such systems.

arXiv:2407.18158

Unlocking Tokens as Data Points for Generalization Bounds on Larger Language Models

Large language models (LLMs) with billions of parameters excel at predicting the next token in a sequence. Recent work computes non-vacuous compression-based generalization bounds for LLMs, but these bounds are vacuous for large models at the billion-parameter scale. Moreover, these bounds are obtained through restrictive compression techniques, bounding compressed models that generate low-quality text. Additionally, the tightness of these existing bounds depends on the number of IID documents in a training set rather than the much larger number of non-IID constituent tokens, leaving untapped potential for tighter bounds. In this work, we instead use properties of martingales to derive generalization bounds that benefit from the vast number of tokens in LLM training sets. Since a dataset contains far more tokens than documents, our generalization bounds not only tolerate but actually benefit from far less restrictive compression schemes. With Monarch matrices, Kronecker factorizations, and post-training quantization, we achieve non-vacuous generalization bounds for LLMs as large as LLaMA2-70B. Unlike previous approaches, our work achieves the first non-vacuous bounds for models that are deployed in practice and generate high-quality text.

arXiv:2407.17200

Generalization Bounds of Surrogate Policies for Combinatorial Optimization Problems

A recent stream of structured learning approaches has improved the practical state of the art for a range of combinatorial optimization problems with complex objectives encountered in operations research. Such approaches train policies that chain a statistical model with a surrogate combinatorial optimization oracle to map any instance of the problem to a feasible solution. The key idea is to exploit the statistical distribution over instances instead of dealing with instances separately. However learning such policies by risk minimization is challenging because the empirical risk is piecewise constant in the parameters, and few theoretical guarantees have been provided so far. In this article, we investigate methods that smooth the risk by perturbing the policy, which eases optimization and improves generalization. Our main contribution is a generalization bound that controls the perturbation bias, the statistical learning error, and the optimization error. Our analysis relies on the introduction of a uniform weak property, which captures and quantifies the interplay of the statistical model and the surrogate combinatorial optimization oracle. This property holds under mild assumptions on the statistical model, the surrogate optimization, and the instance data distribution. We illustrate the result on a range of applications such as stochastic vehicle scheduling. In particular, such policies are relevant for contextual stochastic optimization and our results cover this case.

arXiv:2407.08723

Topological Generalization Bounds for Discrete-Time Stochastic Optimization Algorithms

We present a novel set of rigorous and computationally efficient topology-based complexity notions that exhibit a strong correlation with the generalization gap in modern deep neural networks (DNNs). DNNs show remarkable generalization properties, yet the source of these capabilities remains elusive, defying the established statistical learning theory. Recent studies have revealed that properties of training trajectories can be indicative of generalization. Building on this insight, state-of-the-art methods have leveraged the topology of these trajectories, particularly their fractal dimension, to quantify generalization. Most existing works compute this quantity by assuming continuous- or infinite-time training dynamics, complicating the development of practical estimators capable of accurately predicting generalization without access to test data. In this paper, we respect the discrete-time nature of training trajectories and investigate the underlying topological quantities that can be amenable to topological data analysis tools. This leads to a new family of reliable topological complexity measures that provably bound the generalization error, eliminating the need for restrictive geometric assumptions. These measures are computationally friendly, enabling us to propose simple yet effective algorithms for computing generalization indices. Moreover, our flexible framework can be extended to different domains, tasks, and architectures. Our experimental results demonstrate that our new complexity measures correlate highly with generalization error in industry-standards architectures such as transformers and deep graph networks. Our approach consistently outperforms existing topological bounds across a wide range of datasets, models, and optimizers, highlighting the practical relevance and effectiveness of our complexity measures.

arXiv:2406.16608

When Invariant Representation Learning Meets Label Shift: Insufficiency and Theoretical Insights

As a crucial step toward real-world learning scenarios with changing environments, dataset shift theory and invariant representation learning algorithm have been extensively studied to relax the identical distribution assumption in classical learning setting. Among the different assumptions on the essential of shifting distributions, generalized label shift (GLS) is the latest developed one which shows great potential to deal with the complex factors within the shift. In this paper, we aim to explore the limitations of current dataset shift theory and algorithm, and further provide new insights by presenting a comprehensive understanding of GLS. From theoretical aspect, two informative generalization bounds are derived, and the GLS learner is proved to be sufficiently close to optimal target model from the Bayesian perspective. The main results show the insufficiency of invariant representation learning, and prove the sufficiency and necessity of GLS correction for generalization, which provide theoretical supports and innovations for exploring generalizable model under dataset shift. From methodological aspect, we provide a unified view of existing shift correction frameworks, and propose a kernel embedding-based correction algorithm (KECA) to minimize the generalization error and achieve successful knowledge transfer. Both theoretical results and extensive experiment evaluations demonstrate the sufficiency and necessity of GLS correction for addressing dataset shift and the superiority of proposed algorithm.

arXiv:2406.13928

Optimal deep learning of holomorphic operators between Banach spaces

Operator learning problems arise in many key areas of scientific computing where Partial Differential Equations (PDEs) are used to model physical systems. In such scenarios, the operators map between Banach or Hilbert spaces. In this work, we tackle the problem of learning operators between Banach spaces, in contrast to the vast majority of past works considering only Hilbert spaces. We focus on learning holomorphic operators - an important class of problems with many applications. We combine arbitrary approximate encoders and decoders with standard feedforward Deep Neural Network (DNN) architectures - specifically, those with constant width exceeding the depth - under standard $\ell^2$-loss minimization. We first identify a family of DNNs such that the resulting Deep Learning (DL) procedure achieves optimal generalization bounds for such operators. For standard fully-connected architectures, we then show that there are uncountably many minimizers of the training problem that yield equivalent optimal performance. The DNN architectures we consider are `problem agnostic', with width and depth only depending on the amount of training data $m$ and not on regularity assumptions of the target operator. Next, we show that DL is optimal for this problem: no recovery procedure can surpass these generalization bounds up to log terms. Finally, we present numerical results demonstrating the practical performance on challenging problems including the parametric diffusion, Navier-Stokes-Brinkman and Boussinesq PDEs.