Many machine learning problems can be formulated as minimax problems such as Generative Adversarial Networks (GANs), AUC maximization and robust estimation, to mention but a few. A substantial amount of studies are devoted to studying the convergence behavior of their stochastic gradient-type algorithms. In contrast, there is relatively little work on their generalization, i.e., how the learning models built from training examples would behave on test examples. In this paper, we provide a comprehensive generalization analysis of stochastic gradient methods for minimax problems under both convex-concave and nonconvex-nonconcave cases through the lens of algorithmic stability. We establish a quantitative connection between stability and several generalization measures both in expectation and with high probability. For the convex-concave setting, our stability analysis shows that stochastic gradient descent ascent attains optimal generalization bounds for both smooth and nonsmooth minimax problems. We also establish generalization bounds for both weakly-convex-weakly-concave and gradient-dominated problems.
相關內容
In this work, we study a new recursive stochastic algorithm for the joint estimation of quantile and superquantile of an unknown distribution. The novelty of this algorithm is to use the Cesaro averaging of the quantile estimation inside the recursive approximation of the superquantile. We provide some sharp non-asymptotic bounds on the quadratic risk of the superquantile estimator for different step size sequences. We also prove new non-asymptotic $L^p$-controls on the Robbins Monro algorithm for quantile estimation and its averaged version. Finally, we derive a central limit theorem of our joint procedure using the diffusion approximation point of view hidden behind our stochastic algorithm.
Motivated by the high-frequency data streams continuously generated, real-time learning is becoming increasingly important. These data streams should be processed sequentially with the property that the stream may change over time. In this streaming setting, we propose techniques for minimizing a convex objective through unbiased estimates of its gradients, commonly referred to as stochastic approximation problems. Our methods rely on stochastic approximation algorithms due to their computationally advantage as they only use the previous iterate as a parameter estimate. The reasoning includes iterate averaging that guarantees optimal statistical efficiency under classical conditions. Our non-asymptotic analysis shows accelerated convergence by selecting the learning rate according to the expected data streams. We show that the average estimate converges optimally and robustly to any data stream rate. In addition, noise reduction can be achieved by processing the data in a specific pattern, which is advantageous for large-scale machine learning. These theoretical results are illustrated for various data streams, showing the effectiveness of the proposed algorithms.
Stochastic gradient descent (SGD) is one of the most popular algorithms in modern machine learning. The noise encountered in these applications is different from that in many theoretical analyses of stochastic gradient algorithms. In this article, we discuss some of the common properties of energy landscapes and stochastic noise encountered in machine learning problems, and how they affect SGD-based optimization. In particular, we show that the learning rate in SGD with machine learning noise can be chosen to be small, but uniformly positive for all times if the energy landscape resembles that of overparametrized deep learning problems. If the objective function satisfies a Lojasiewicz inequality, SGD converges to the global minimum exponentially fast, and even for functions which may have local minima, we establish almost sure convergence to the global minimum at an exponential rate from any finite energy initialization. The assumptions that we make in this result concern the behavior where the objective function is either small or large and the nature of the gradient noise, but the energy landscape is fairly unconstrained on the domain where the objective function takes values in an intermediate regime.
In this work, we study an inverse problem of recovering a space-time dependent diffusion coefficient in the subdiffusion model from the distributed observation, where the mathematical model involves a Djrbashian-Caputo fractional derivative of order $\alpha\in(0,1)$ in time. The main technical challenges of both theoretical and numerical analysis lie in the limited smoothing properties due to the fractional differential operator and the high degree of nonlinearity of the forward map from the unknown diffusion coefficient to the distributed observation. Theoretically, we establish two conditional stability results using a novel test function, which leads to a stability bound in $L^2(0,T;L^2(\Omega))$ under a suitable positivity condition. The positivity condition is verified for a large class of problem data. Numerically, we develop a rigorous procedure for the recovery of the diffusion coefficient based on a regularized least-squares formulation, which is then discretized by the standard Galerkin method with continuous piecewise linear elements in space and backward Euler convolution quadrature in time. We provide a complete error analysis of the fully discrete formulation, by combining several new error estimates for the direct problem (optimal in terms of data regularity), a discrete version of fractional maximal $L^p$ regularity, and a nonstandard energy argument. Under the positivity condition, we obtain a standard $L^2(0,T; L^2(\Omega))$ error estimate consistent with the conditional stability. Further, we illustrate the analysis with some numerical examples.
In the paper, we propose a class of faster adaptive Gradient Descent Ascent (GDA) methods for solving the nonconvex-strongly-concave minimax problems based on unified adaptive matrices, which include almost existing coordinate-wise and global adaptive learning rates. Specifically, we propose a fast Adaptive Gradient Decent Ascent (AdaGDA) method based on the basic momentum technique, which reaches a lower sample complexity of $O(\kappa^4\epsilon^{-4})$ for finding an $\epsilon$-stationary point without large batches, which improves the results of the existing adaptive GDA methods by a factor of $O(\sqrt{\kappa})$. At the same time, we present an accelerated version of AdaGDA (VR-AdaGDA) method based on the momentum-based variance reduced technique, which achieves a lower sample complexity of $O(\kappa^{4.5}\epsilon^{-3})$ for finding an $\epsilon$-stationary point without large batches, which improves the results of the existing adaptive GDA methods by a factor of $O(\epsilon^{-1})$. Moreover, we prove that our VR-AdaGDA method reaches the best known sample complexity of $O(\kappa^{3}\epsilon^{-3})$ with the mini-batch size $O(\kappa^3)$. In particular, we provide an effective convergence analysis framework for our adaptive GDA methods. Some experimental results on fair classifier and policy evaluation tasks demonstrate the efficiency of our algorithms.
Chance constrained optimization problems allow to model problems where constraints involving stochastic components should only be violated with a small probability. Evolutionary algorithms have recently been applied to this scenario and shown to achieve high quality results. With this paper, we contribute to the theoretical understanding of evolutionary algorithms for chance constrained optimization. We study the scenario of stochastic components that are independent and Normally distributed. By generalizing results for the class of linear functions to the sum of transformed linear functions, we show that the (1+1)~EA can optimize the chance constrained setting without additional constraints in time O(n log n). However, we show that imposing an additional uniform constraint already leads to local optima for very restricted scenarios and an exponential optimization time for the (1+1)~EA. We therefore propose a multi-objective formulation of the problem which trades off the expected cost and its variance. We show that multi-objective evolutionary algorithms are highly effective when using this formulation and obtain a set of solutions that contains an optimal solution for any possible confidence level imposed on the constraint. Furthermore, we show that this approach can also be used to compute a set of optimal solutions for the chance constrained minimum spanning tree problem.
In recent years, distributed optimization is proven to be an effective approach to accelerate training of large scale machine learning models such as deep neural networks. With the increasing computation power of GPUs, the bottleneck of training speed in distributed training is gradually shifting from computation to communication. Meanwhile, in the hope of training machine learning models on mobile devices, a new distributed training paradigm called ``federated learning'' has become popular. The communication time in federated learning is especially important due to the low bandwidth of mobile devices. While various approaches to improve the communication efficiency have been proposed for federated learning, most of them are designed with SGD as the prototype training algorithm. While adaptive gradient methods have been proven effective for training neural nets, the study of adaptive gradient methods in federated learning is scarce. In this paper, we propose an adaptive gradient method that can guarantee both the convergence and the communication efficiency for federated learning.
Non-convex optimization is ubiquitous in modern machine learning. Researchers devise non-convex objective functions and optimize them using off-the-shelf optimizers such as stochastic gradient descent and its variants, which leverage the local geometry and update iteratively. Even though solving non-convex functions is NP-hard in the worst case, the optimization quality in practice is often not an issue -- optimizers are largely believed to find approximate global minima. Researchers hypothesize a unified explanation for this intriguing phenomenon: most of the local minima of the practically-used objectives are approximately global minima. We rigorously formalize it for concrete instances of machine learning problems.
Self-training algorithms, which train a model to fit pseudolabels predicted by another previously-learned model, have been very successful for learning with unlabeled data using neural networks. However, the current theoretical understanding of self-training only applies to linear models. This work provides a unified theoretical analysis of self-training with deep networks for semi-supervised learning, unsupervised domain adaptation, and unsupervised learning. At the core of our analysis is a simple but realistic ``expansion'' assumption, which states that a low-probability subset of the data must expand to a neighborhood with large probability relative to the subset. We also assume that neighborhoods of examples in different classes have minimal overlap. We prove that under these assumptions, the minimizers of population objectives based on self-training and input-consistency regularization will achieve high accuracy with respect to ground-truth labels. By using off-the-shelf generalization bounds, we immediately convert this result to sample complexity guarantees for neural nets that are polynomial in the margin and Lipschitzness. Our results help explain the empirical successes of recently proposed self-training algorithms which use input consistency regularization.
Asynchronous momentum stochastic gradient descent algorithms (Async-MSGD) is one of the most popular algorithms in distributed machine learning. However, its convergence properties for these complicated nonconvex problems is still largely unknown, because of the current technical limit. Therefore, in this paper, we propose to analyze the algorithm through a simpler but nontrivial nonconvex problem - streaming PCA, which helps us to understand Aync-MSGD better even for more general problems. Specifically, we establish the asymptotic rate of convergence of Async-MSGD for streaming PCA by diffusion approximation. Our results indicate a fundamental tradeoff between asynchrony and momentum: To ensure convergence and acceleration through asynchrony, we have to reduce the momentum (compared with Sync-MSGD). To the best of our knowledge, this is the first theoretical attempt on understanding Async-MSGD for distributed nonconvex stochastic optimization. Numerical experiments on both streaming PCA and training deep neural networks are provided to support our findings for Async-MSGD.
小貼士
登錄享
相關主題
注冊地址： 北京市海淀區羊坊店路18號2幢3層301-191