Lower Bounds and Accelerated Algorithms in Distributed Stochastic Optimization with Communication Compression

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: He, Yutong, Huang, Xinmeng, Chen, Yiming, Yin, Wotao, Yuan, Kun
Format: Preprint
Published: 2023
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866913743059812352
author He, Yutong
Huang, Xinmeng
Chen, Yiming
Yin, Wotao
Yuan, Kun
author_facet He, Yutong
Huang, Xinmeng
Chen, Yiming
Yin, Wotao
Yuan, Kun
contents Communication compression is an essential strategy for alleviating communication overhead by reducing the volume of information exchanged between computing nodes in large-scale distributed stochastic optimization. Although numerous algorithms with convergence guarantees have been obtained, the optimal performance limit under communication compression remains unclear. In this paper, we investigate the performance limit of distributed stochastic optimization algorithms employing communication compression. We focus on two main types of compressors, unbiased and contractive, and address the best-possible convergence rates one can obtain with these compressors. We establish the lower bounds for the convergence rates of distributed stochastic optimization in six different settings, combining strongly-convex, generally-convex, or non-convex functions with unbiased or contractive compressor types. To bridge the gap between lower bounds and existing algorithms' rates, we propose NEOLITHIC, a nearly optimal algorithm with compression that achieves the established lower bounds up to logarithmic factors under mild conditions. Extensive experimental results support our theoretical findings. This work provides insights into the theoretical limitations of existing compressors and motivates further research into fundamentally new compressor properties.
format Preprint
id arxiv_https___arxiv_org_abs_2305_07612
institution arXiv
publishDate 2023
record_format arxiv
spellingShingle Lower Bounds and Accelerated Algorithms in Distributed Stochastic Optimization with Communication Compression
He, Yutong
Huang, Xinmeng
Chen, Yiming
Yin, Wotao
Yuan, Kun
Machine Learning
Distributed, Parallel, and Cluster Computing
Optimization and Control
Communication compression is an essential strategy for alleviating communication overhead by reducing the volume of information exchanged between computing nodes in large-scale distributed stochastic optimization. Although numerous algorithms with convergence guarantees have been obtained, the optimal performance limit under communication compression remains unclear. In this paper, we investigate the performance limit of distributed stochastic optimization algorithms employing communication compression. We focus on two main types of compressors, unbiased and contractive, and address the best-possible convergence rates one can obtain with these compressors. We establish the lower bounds for the convergence rates of distributed stochastic optimization in six different settings, combining strongly-convex, generally-convex, or non-convex functions with unbiased or contractive compressor types. To bridge the gap between lower bounds and existing algorithms' rates, we propose NEOLITHIC, a nearly optimal algorithm with compression that achieves the established lower bounds up to logarithmic factors under mild conditions. Extensive experimental results support our theoretical findings. This work provides insights into the theoretical limitations of existing compressors and motivates further research into fundamentally new compressor properties.
title Lower Bounds and Accelerated Algorithms in Distributed Stochastic Optimization with Communication Compression
topic Machine Learning
Distributed, Parallel, and Cluster Computing
Optimization and Control
url https://arxiv.org/abs/2305.07612