From PowerSGD to PowerSGD+: Low-Rank Gradient Compression for Distributed Optimization with Convergence Guarantees

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Xie, Shengping, Chen, Chuyan, Yuan, Kun
Format: Preprint
Published: 2025
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866912585939419136
author Xie, Shengping
Chen, Chuyan
Yuan, Kun
author_facet Xie, Shengping
Chen, Chuyan
Yuan, Kun
contents Low-rank gradient compression methods, such as PowerSGD, have gained attention in communication-efficient distributed optimization. However, the convergence guarantees of PowerSGD remain unclear, particularly in stochastic settings. In this paper, we show that PowerSGD does not always converge to the optimal solution and provide a clear counterexample to support this finding. To address this, we introduce PowerSGD+, which periodically updates the projection subspace via singular value decomposition, ensuring that it remains aligned with the optimal subspace. We prove that PowerSGD+ converges under standard assumptions and validate its effectiveness through empirical evaluation on large language model tasks.
format Preprint
id arxiv_https___arxiv_org_abs_2509_11254
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle From PowerSGD to PowerSGD+: Low-Rank Gradient Compression for Distributed Optimization with Convergence Guarantees
Xie, Shengping
Chen, Chuyan
Yuan, Kun
Optimization and Control
Machine Learning
Low-rank gradient compression methods, such as PowerSGD, have gained attention in communication-efficient distributed optimization. However, the convergence guarantees of PowerSGD remain unclear, particularly in stochastic settings. In this paper, we show that PowerSGD does not always converge to the optimal solution and provide a clear counterexample to support this finding. To address this, we introduce PowerSGD+, which periodically updates the projection subspace via singular value decomposition, ensuring that it remains aligned with the optimal subspace. We prove that PowerSGD+ converges under standard assumptions and validate its effectiveness through empirical evaluation on large language model tasks.
title From PowerSGD to PowerSGD+: Low-Rank Gradient Compression for Distributed Optimization with Convergence Guarantees
topic Optimization and Control
Machine Learning
url https://arxiv.org/abs/2509.11254