Constant Stepsize Local GD for Logistic Regression: Acceleration by Instability

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Crawshaw, Michael, Woodworth, Blake, Liu, Mingrui
Format: Preprint
Published: 2025
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866912433659969536
author Crawshaw, Michael
Woodworth, Blake
Liu, Mingrui
author_facet Crawshaw, Michael
Woodworth, Blake
Liu, Mingrui
contents Existing analysis of Local (Stochastic) Gradient Descent for heterogeneous objectives requires stepsizes $η\leq 1/K$ where $K$ is the communication interval, which ensures monotonic decrease of the objective. In contrast, we analyze Local Gradient Descent for logistic regression with separable, heterogeneous data using any stepsize $η> 0$. With $R$ communication rounds and $M$ clients, we show convergence at a rate $\mathcal{O}(1/ηK R)$ after an initial unstable phase lasting for $\widetilde{\mathcal{O}}(ηK M)$ rounds. This improves upon the existing $\mathcal{O}(1/R)$ rate for general smooth, convex objectives. Our analysis parallels the single machine analysis of~\cite{wu2024large} in which instability is caused by extremely large stepsizes, but in our setting another source of instability is large local updates with heterogeneous objectives.
format Preprint
id arxiv_https___arxiv_org_abs_2506_13974
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Constant Stepsize Local GD for Logistic Regression: Acceleration by Instability
Crawshaw, Michael
Woodworth, Blake
Liu, Mingrui
Machine Learning
Existing analysis of Local (Stochastic) Gradient Descent for heterogeneous objectives requires stepsizes $η\leq 1/K$ where $K$ is the communication interval, which ensures monotonic decrease of the objective. In contrast, we analyze Local Gradient Descent for logistic regression with separable, heterogeneous data using any stepsize $η> 0$. With $R$ communication rounds and $M$ clients, we show convergence at a rate $\mathcal{O}(1/ηK R)$ after an initial unstable phase lasting for $\widetilde{\mathcal{O}}(ηK M)$ rounds. This improves upon the existing $\mathcal{O}(1/R)$ rate for general smooth, convex objectives. Our analysis parallels the single machine analysis of~\cite{wu2024large} in which instability is caused by extremely large stepsizes, but in our setting another source of instability is large local updates with heterogeneous objectives.
title Constant Stepsize Local GD for Logistic Regression: Acceleration by Instability
topic Machine Learning
url https://arxiv.org/abs/2506.13974