Error Lower Bounds of Constant Step-size Stochastic Gradient Descent
Optimization and Control
2019-10-21 v1 Machine Learning
Abstract
Stochastic Gradient Descent (SGD) plays a central role in modern machine learning. While there is extensive work on providing error upper bound for SGD, not much is known about SGD error lower bound. In this paper, we study the convergence of constant step-size SGD. We provide error lower bound of SGD for potentially non-convex objective functions with Lipschitz gradients. To our knowledge, this is the first analysis for SGD error lower bound without the strong convexity assumption. We use experiments to illustrate our theoretical results.
Cite
@article{arxiv.1910.08212,
title = {Error Lower Bounds of Constant Step-size Stochastic Gradient Descent},
author = {Zhiyan Ding and Yiding Chen and Qin Li and Xiaojin Zhu},
journal= {arXiv preprint arXiv:1910.08212},
year = {2019}
}