On the Periodic Behavior of Neural Network Training with Batch Normalization and Weight Decay

Abstract

Training neural networks with batch normalization and weight decay has become a common practice in recent years. In this work, we show that their combined use may result in a surprising periodic behavior of optimization dynamics: the training process regularly exhibits destabilizations that, however, do not lead to complete divergence but cause a new period of training. We rigorously investigate the mechanism underlying the discovered periodic behavior from both empirical and theoretical points of view and analyze the conditions in which it occurs in practice. We also demonstrate that periodic behavior can be regarded as a generalization of two previously opposing perspectives on training with batch normalization and weight decay, namely the equilibrium presumption and the instability presumption.

Published in NeurIPS 2021. First two authors contributed equally

References in corpus (12)

• Batch Normalization: Accelerating Deep Network Training by Reducing Internal Covariate Shift
• Instance Normalization: The Missing Ingredient for Fast Stylization
• Deep Ensembles: A Loss Landscape Perspective
• L2 Regularization versus Batch and Weight Normalization
• An Investigation into Neural Net Optimization via Hessian Eigenvalue Density
• Norm matters: efficient and accurate normalization schemes in deep networks
• Theoretical Analysis of Auto Rate-Tuning by Batch Normalization
• A Mean Field Theory of Batch Normalization
• Online Normalization for Training Neural Networks
• Gradient Descent on Neural Networks Typically Occurs at the Edge of Stability
• Reconciling Modern Deep Learning with Traditional Optimization Analyses: The Intrinsic Learning Rate
• Spherical Motion Dynamics: Learning Dynamics of Neural Network with Normalization, Weight Decay, and SGD