Stochastic Collapse: How Gradient Noise Attracts SGD Dynamics Towards Simpler Subnetworks

• URL: http://arxiv.org/abs/2306.04251v3
• Date: Wed, 29 May 2024 01:03:31 GMT
• Title: Stochastic Collapse: How Gradient Noise Attracts SGD Dynamics Towards Simpler Subnetworks
• Authors: Feng Chen, Daniel Kunin, Atsushi Yamamura, Surya Ganguli,
• Abstract summary: We reveal a strong implicit bias of gradient of descent (SGD) that drives overly expressive networks to much simplerworks. We focus on two classes of invariant sets that correspond to simpler (sparse or low-rank)works and commonly appear in modern architectures. We observe empirically the existence of attractive invariant sets in trained deep neural networks, implying that SGD dynamics often collapses vanishing simpleworks with either redundant neurons.
• Score: 28.87871359825978
• License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/
• Abstract: In this work, we reveal a strong implicit bias of stochastic gradient descent (SGD) that drives overly expressive networks to much simpler subnetworks, thereby dramatically reducing the number of independent parameters, and improving generalization. To reveal this bias, we identify invariant sets, or subsets of parameter space that remain unmodified by SGD. We focus on two classes of invariant sets that correspond to simpler (sparse or low-rank) subnetworks and commonly appear in modern architectures. Our analysis uncovers that SGD exhibits a property of stochastic attractivity towards these simpler invariant sets. We establish a sufficient condition for stochastic attractivity based on a competition between the loss landscape's curvature around the invariant set and the noise introduced by stochastic gradients. Remarkably, we find that an increased level of noise strengthens attractivity, leading to the emergence of attractive invariant sets associated with saddle-points or local maxima of the train loss. We observe empirically the existence of attractive invariant sets in trained deep neural networks, implying that SGD dynamics often collapses to simple subnetworks with either vanishing or redundant neurons. We further demonstrate how this simplifying process of stochastic collapse benefits generalization in a linear teacher-student framework. Finally, through this analysis, we mechanistically explain why early training with large learning rates for extended periods benefits subsequent generalization.

Related papers

• On the Dynamics Under the Unhinged Loss and Beyond [104.49565602940699]
  We introduce the unhinged loss, a concise loss function, that offers more mathematical opportunities to analyze closed-form dynamics. The unhinged loss allows for considering more practical techniques, such as time-vary learning rates and feature normalization.
  arXiv  Detail & Related papers  (2023-12-13T02:11:07Z)
• Machine learning in and out of equilibrium [58.88325379746631]
  Our study uses a Fokker-Planck approach, adapted from statistical physics, to explore these parallels. We focus in particular on the stationary state of the system in the long-time limit, which in conventional SGD is out of equilibrium. We propose a new variation of Langevin dynamics (SGLD) that harnesses without replacement minibatching.
  arXiv  Detail & Related papers  (2023-06-06T09:12:49Z)
• Escaping mediocrity: how two-layer networks learn hard generalized linear models with SGD [29.162265194920522]
  This study explores the sample complexity for two-layer neural networks to learn a generalized linear target function under Gradient Descent (SGD) We show that overfactorization can only enhance convergence by a constant factor within this problem class. Yet, we demonstrate that a deterministic approximation of this process adequately represents the escape time, implying that the role of SGDity may be minimal in this scenario.
  arXiv  Detail & Related papers  (2023-05-29T14:40:56Z)
• Implicit Regularization for Group Sparsity [33.487964460794764]
  We show that gradient descent over the squared regression loss, without any explicit regularization, biases towards solutions with a group sparsity structure. We analyze the gradient dynamics of the corresponding regression problem in the general noise setting and obtain minimax-optimal error rates. In the degenerate case of size-one groups, our approach gives rise to a new algorithm for sparse linear regression.
  arXiv  Detail & Related papers  (2023-01-29T20:54:03Z)
• SGD with Large Step Sizes Learns Sparse Features [22.959258640051342]
  We showcase important features of the dynamics of the Gradient Descent (SGD) in the training of neural networks. We show that the longer large step sizes keep SGD high in the loss landscape, the better the implicit regularization can operate and find sparse representations.
  arXiv  Detail & Related papers  (2022-10-11T11:00:04Z)
• Learning an Invertible Output Mapping Can Mitigate Simplicity Bias in Neural Networks [66.76034024335833]
  We investigate why diverse/ complex features are learned by the backbone, and their brittleness is due to the linear classification head relying primarily on the simplest features. We propose Feature Reconstruction Regularizer (FRR) to ensure that the learned features can be reconstructed back from the logits. We demonstrate up to 15% gains in OOD accuracy on the recently introduced semi-synthetic datasets with extreme distribution shifts.
  arXiv  Detail & Related papers  (2022-10-04T04:01:15Z)
• Clipped Stochastic Methods for Variational Inequalities with Heavy-Tailed Noise [64.85879194013407]
  We prove the first high-probability results with logarithmic dependence on the confidence level for methods for solving monotone and structured non-monotone VIPs. Our results match the best-known ones in the light-tails case and are novel for structured non-monotone problems. In addition, we numerically validate that the gradient noise of many practical formulations is heavy-tailed and show that clipping improves the performance of SEG/SGDA.
  arXiv  Detail & Related papers  (2022-06-02T15:21:55Z)
• Stochastic Training is Not Necessary for Generalization [57.04880404584737]
  It is widely believed that the implicit regularization of gradient descent (SGD) is fundamental to the impressive generalization behavior we observe in neural networks. In this work, we demonstrate that non-stochastic full-batch training can achieve strong performance on CIFAR-10 that is on-par with SGD.
  arXiv  Detail & Related papers  (2021-09-29T00:50:00Z)
• Shallow Univariate ReLu Networks as Splines: Initialization, Loss Surface, Hessian, & Gradient Flow Dynamics [1.5393457051344297]
  We propose reparametrizing ReLU NNs as continuous piecewise linear splines. We develop a surprisingly simple and transparent view of the structure of the loss surface, including its critical and fixed points, Hessian, and Hessian spectrum. Videos of learning dynamics using a spline-based visualization are available at http://shorturl.at/tFWZ2.
  arXiv  Detail & Related papers  (2020-08-04T19:19:49Z)
• The Heavy-Tail Phenomenon in SGD [7.366405857677226]
  We show that depending on the structure of the Hessian of the loss at the minimum, the SGD iterates will converge to a emphheavy-tailed stationary distribution. We translate our results into insights about the behavior of SGD in deep learning.
  arXiv  Detail & Related papers  (2020-06-08T16:43:56Z)
• Kernel and Rich Regimes in Overparametrized Models [69.40899443842443]
  We show that gradient descent on overparametrized multilayer networks can induce rich implicit biases that are not RKHS norms. We also demonstrate this transition empirically for more complex matrix factorization models and multilayer non-linear networks.
  arXiv  Detail & Related papers  (2020-02-20T15:43:02Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.