Straggler management

Syed Zawad, Feng Yan, Ali Anwar

Research output: Chapter in Book/Report/Conference proceedingChapter

Abstract

For this chapter, we elaborate on one of the most common challenge in Federated Learning-stragglers. The chapters "Local Training and Scalability of Federated Learning Systems" and "Introduction to Federated Learning Systems" have talked briefly about it, and we delve even deeper here. We first provide an introduction on what the problem is and why it is important. We talk about a study to show the effect of stragglers in a practical setting. As an example, we then talk about TiFL, a framework that proposes to solve such a problem using grouping. Empirical results are presented to show how such systems may help mitigate the effect of stragglers.

Original languageEnglish (US)
Title of host publicationFederated Learning
Subtitle of host publicationA Comprehensive Overview of Methods and Applications
PublisherSpringer International Publishing
Pages235-258
Number of pages24
ISBN (Print)9783030968960
DOIs
StatePublished - Jul 7 2022
Externally publishedYes

Bibliographical note

Publisher Copyright:
© The Author(s), under exclusive license to Springer Nature Switzerland AG 2022.

Fingerprint

Dive into the research topics of 'Straggler management'. Together they form a unique fingerprint.

Cite this