Generalization of Silver Stepsize Schedule to Stochastic Optimization

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Bai, Luwei, Zeng, Yang, Zhou, Baoyu
Format: Preprint
Published: 2025
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866915640306040832
author Bai, Luwei
Zeng, Yang
Zhou, Baoyu
author_facet Bai, Luwei
Zeng, Yang
Zhou, Baoyu
contents This work introduces a two-step stepsize schedule for stochastic gradient methods minimizing smooth strongly convex functions. We consider the setting where only stochastic gradient approximations, which are unbiased, of bounded variance, and supported on a finite set, are accessible. When the variance bound is relatively smaller than a ratio of the initial optimality gap, the proposed stepsize schedule achieves better convergence performance compared to the well-regarded constant stepsize α = 2/(M+m), where m and M denote the strong convexity and gradient-Lipschitz parameters, respectively. Our stepsize schedule can be viewed as a generalization of the well-known two-step silver stepsize schedule in [J. M. Altschuler and P. A. Parrilo, Journal of the ACM, 72(2):1-38, 2025] from deterministic setting to stochastic optimization.
format Preprint
id arxiv_https___arxiv_org_abs_2511_21917
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Generalization of Silver Stepsize Schedule to Stochastic Optimization
Bai, Luwei
Zeng, Yang
Zhou, Baoyu
Optimization and Control
This work introduces a two-step stepsize schedule for stochastic gradient methods minimizing smooth strongly convex functions. We consider the setting where only stochastic gradient approximations, which are unbiased, of bounded variance, and supported on a finite set, are accessible. When the variance bound is relatively smaller than a ratio of the initial optimality gap, the proposed stepsize schedule achieves better convergence performance compared to the well-regarded constant stepsize α = 2/(M+m), where m and M denote the strong convexity and gradient-Lipschitz parameters, respectively. Our stepsize schedule can be viewed as a generalization of the well-known two-step silver stepsize schedule in [J. M. Altschuler and P. A. Parrilo, Journal of the ACM, 72(2):1-38, 2025] from deterministic setting to stochastic optimization.
title Generalization of Silver Stepsize Schedule to Stochastic Optimization
topic Optimization and Control
url https://arxiv.org/abs/2511.21917