1. A Diffusion Approximation Theory of Momentum Stochastic Gradient Descent in Nonconvex Optimization
- Author
-
Zhehui Chen, Tuo Zhao, Enlu Zhou, and Tianyi Liu
- Subjects
Statistics and Probability ,Momentum (technical analysis) ,Stochastic gradient descent ,Optimization problem ,Computer science ,Modeling and Simulation ,Bayesian probability ,Applied mathematics ,Deep neural networks ,Management Science and Operations Research ,Statistics, Probability and Uncertainty ,Heavy traffic approximation - Abstract
Momentum stochastic gradient descent (MSGD) algorithm has been widely applied to many nonconvex optimization problems in machine learning (e.g., training deep neural networks, variational Bayesian inference, etc.). Despite its empirical success, there is still a lack of theoretical understanding of convergence properties of MSGD. To fill this gap, we propose to analyze the algorithmic behavior of MSGD by diffusion approximations for nonconvex optimization problems with strict saddle points and isolated local optima. Our study shows that the momentum helps escape from saddle points but hurts the convergence within the neighborhood of optima (if without the step size annealing or momentum annealing). Our theoretical discovery partially corroborates the empirical success of MSGD in training deep neural networks.
- Published
- 2021