LGJun 16, 2025

Constant Stepsize Local GD for Logistic Regression: Acceleration by Instability

arXiv:2506.13974v12 citationsh-index: 8ICML
Originality Highly original
AI Analysis

This addresses the challenge of communication efficiency in federated learning for logistic regression, offering a faster convergence method with theoretical guarantees, though it is incremental in extending single-machine analysis to a distributed setting.

The paper tackles the problem of Local Gradient Descent for logistic regression with heterogeneous data by allowing any stepsize, showing convergence at a rate of O(1/ηKR) after an initial unstable phase of O(ηKM) rounds, which improves upon the existing O(1/R) rate for general convex objectives.

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.

Foundations

The foundational work for this paper's niche, ranked by how specifically the neighbourhood builds on it — not by global fame.

Your Notes