Analyzing the convergence of federated learning with biased client participation

Lei Tan, Miao Hu, Yipeng Zhou, Di Wu*

*Corresponding author for this work

Research output: Chapter in Book/Report/Conference proceedingConference proceeding contributionpeer-review

Abstract

Federated Learning (FL) is a promising decentralized machine learning framework that enables a massive number of clients (e.g., smartphones) to collaboratively train a global model over the Internet without sacrificing their privacy. Though FL’s efficacy in non-convex problems is proven, its convergence amidst biased client participation lacks theoretical study. In this paper, we analyze the convergence of FedAvg on non-convex problems, which is the most renowned FL algorithm. We assume even data distribution but non-IID among clients, and elucidate the convergence rate of FedAvg in situations characterized by biased client participation. Our analysis reveals that biased client participation can significantly reduce the precision of the FL model. We validate this through trace-driven experiments, demonstrating that unbiased client participation results in 11% to 50% higher test accuracy compared to extremely biased client participation.

Original languageEnglish
Title of host publicationAdvanced Data Mining and Applications
Subtitle of host publication19th International Conference, ADMA 2023, Shenyang, China, August 21–23, 2023, proceedings, part II
EditorsXiaochun Yang, Heru Suhartanto, Guoren Wang, Bin Wang, Jing Jiang, Bing Li, Huaijie Zhu, Ningning Cui
Place of PublicationCham
PublisherSpringer, Springer Nature
Pages423-439
Number of pages17
ISBN (Electronic)9783031466649
ISBN (Print)9783031466632
DOIs
Publication statusPublished - 2023
Event19th International Conference on Advanced Data Mining and Applications, ADMA 2023 - Shenyang, China
Duration: 21 Aug 202323 Aug 2023

Publication series

NameLecture Notes in Computer Science
PublisherSpringer
Volume14177
ISSN (Print)0302-9743
ISSN (Electronic)1611-3349

Conference

Conference19th International Conference on Advanced Data Mining and Applications, ADMA 2023
Country/TerritoryChina
CityShenyang
Period21/08/2323/08/23

Keywords

  • Federated learning
  • Non-convex
  • Biased participation
  • Convergence analysis

Fingerprint

Dive into the research topics of 'Analyzing the convergence of federated learning with biased client participation'. Together they form a unique fingerprint.

Cite this