On the Training Instability of Shuffling SGD with Batch Normalization

• URL: http://arxiv.org/abs/2302.12444v3
• Date: Mon, 14 Aug 2023 05:22:10 GMT
• Title: On the Training Instability of Shuffling SGD with Batch Normalization
• Abstract summary: Single Shuffle (SS) and Random Reshuffle (RR) interact surprisingly differently in the presence of batch normalization. We show that SS leads to distorted optima in regression and divergence for classification, whereas RR avoids both distortion and divergence.
• Score: 44.28777474091466
• License: http://creativecommons.org/licenses/by/4.0/
• Abstract: We uncover how SGD interacts with batch normalization and can exhibit undesirable training dynamics such as divergence. More precisely, we study how Single Shuffle (SS) and Random Reshuffle (RR) -- two widely used variants of SGD -- interact surprisingly differently in the presence of batch normalization: RR leads to much more stable evolution of training loss than SS. As a concrete example, for regression using a linear network with batch normalization, we prove that SS and RR converge to distinct global optima that are "distorted" away from gradient descent. Thereafter, for classification we characterize conditions under which training divergence for SS and RR can, and cannot occur. We present explicit constructions to show how SS leads to distorted optima in regression and divergence for classification, whereas RR avoids both distortion and divergence. We validate our results by confirming them empirically in realistic settings, and conclude that the separation between SS and RR used with batch normalization is relevant in practice.

Related papers

• Convex-Hull-Neighborhood Smooth Dual Generalization: Controlling Local Correction Propagation in Offline RL [7.803284875836737]
  bootstrapping can amplify estimation errors in offline reinforcement learning.<n>We propose Convex Hull Neighborhood Smooth Dual Generalization, which expresses the Bellman backup as an in-sample value target plus a CHN-local correction.<n> Experiments on Gym-MuJoCo and AntMaze show strong aggregate performance and stable value estimation.
  arXiv  Detail & Related papers  (2026-08-04T04:30:02Z)
• Improving Search Agent with One Line of Code [68.58667107354253]
  Tool-based Agentic Reinforcement Learning (TARL) has emerged as a promising paradigm for training search agents to interact with external tools.<n>We propose textbfSearch textbfAgent textbfPolicy textbfOptimization (textbfSAPO), which stabilizes training via a conditional token-level KL constraint.
  arXiv  Detail & Related papers  (2026-03-10T04:07:39Z)
• Conditional Rank-Rank Regression via Deep Conditional Transformation Models [8.100828624916574]
  Intergenerational mobility quantifies the transmission of socio-economic outcomes from parents to children.<n> Conditional rank-rank regression (CRRR) measures within-group mobility using conditional ranks.<n>We extend CRRR to discrete outcomes via an $$-indexed conditional-rank definition and study sensitivity to $$.<n>For continuous outcomes, we establish a theory for the proposed estimators and verify the validity of exchangeable inference.
  arXiv  Detail & Related papers  (2026-03-07T14:27:09Z)
• VESPO: Variational Sequence-Level Soft Policy Optimization for Stable Off-Policy LLM Training [18.849117699859622]
  Training stability is a central challenge in reinforcement learning for large language models.<n>We propose Variational sEquence-level Soft Policy Optimization (VESPO)<n> Experiments on mathematical reasoning benchmarks show that VESPO maintains stable training under staleness ratios up to 64x and fully asynchronous execution.
  arXiv  Detail & Related papers  (2026-02-11T09:48:08Z)
• 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)
• Function-space regularized R\'enyi divergences [6.221019624345409]
  We propose a new family of regularized R'enyi divergences parametrized by a variational function space. We prove several properties of these new divergences, showing that they interpolate between the classical R'enyi divergences and IPMs. We show that the proposed regularized R'enyi divergences inherit features from IPMs such as the ability to compare distributions that are not absolutely continuous.
  arXiv  Detail & Related papers  (2022-10-10T19:18:04Z)
• Distributionally Robust Models with Parametric Likelihood Ratios [123.05074253513935]
  Three simple ideas allow us to train models with DRO using a broader class of parametric likelihood ratios. We find that models trained with the resulting parametric adversaries are consistently more robust to subpopulation shifts when compared to other DRO approaches.
  arXiv  Detail & Related papers  (2022-04-13T12:43:12Z)
• Explicit Regularization via Regularizer Mirror Descent [32.0512015286512]
  We propose a new method for training deep neural networks (DNNs) with regularization, called regularizer mirror descent (RMD) RMD simultaneously interpolates the training data and minimizes a certain potential function of the weights. Our results suggest that the performance ofRMD is remarkably robust and significantly better than both gradient descent (SGD) and weight decay.
  arXiv  Detail & Related papers  (2022-02-22T10:21:44Z)
• On the Double Descent of Random Features Models Trained with SGD [78.0918823643911]
  We study properties of random features (RF) regression in high dimensions optimized by gradient descent (SGD) We derive precise non-asymptotic error bounds of RF regression under both constant and adaptive step-size SGD setting. We observe the double descent phenomenon both theoretically and empirically.
  arXiv  Detail & Related papers  (2021-10-13T17:47:39Z)
• 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)
• Re-parameterizing VAEs for stability [1.90365714903665]
  We propose a theoretical approach towards the training numerical stability of Variational AutoEncoders (VAE) Our work is motivated by recent studies empowering VAEs to reach state of the art generative results on complex image datasets. We show that by implementing small changes to the way we parameterize the Normal distributions on which they rely, VAEs can securely be trained.
  arXiv  Detail & Related papers  (2021-06-25T16:19:09Z)
• Generalization Error Rates in Kernel Regression: The Crossover from the Noiseless to Noisy Regime [29.731516232010343]
  We consider Kernel Ridge Regression (KRR) under the Gaussian design. We show the existence of a transition in the noisy setting between the noiseless exponents to its noisy values as the sample complexity is increased.
  arXiv  Detail & Related papers  (2021-05-31T14:39:08Z)
• Benign Overfitting of Constant-Stepsize SGD for Linear Regression [122.70478935214128]
  inductive biases are central in preventing overfitting empirically. This work considers this issue in arguably the most basic setting: constant-stepsize SGD for linear regression. We reflect on a number of notable differences between the algorithmic regularization afforded by (unregularized) SGD in comparison to ordinary least squares.
  arXiv  Detail & Related papers  (2021-03-23T17:15:53Z)

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.