A General Theory for Client Sampling in Federated Learning - Archive ouverte HAL Access content directly
Conference Papers Year : 2022

A General Theory for Client Sampling in Federated Learning

Abstract

While client sampling is a central operation of current state-of-the-art federated learning (FL) approaches, the impact of this procedure on the convergence and speed of FL remains under-investigated. In this work, we provide a general theoretical framework to quantify the impact of a client sampling scheme and of the clients heterogeneity on the federated optimization. First, we provide a unified theoretical ground for previously reported sampling schemes experimental results on the relationship between FL convergence and the variance of the aggregation weights. Second, we prove for the first time that the quality of FL convergence is also impacted by the resulting covariance between aggregation weights. Our theory is general, and is here applied to Multinomial Distribution (MD) and Uniform sampling, two default unbiased client sampling schemes of FL, and demonstrated through a series of experiments in non-iid and unbalanced scenarios. Our results suggest that MD sampling should be used as default sampling scheme, due to the resilience to the changes in data ratio during the learning process, while Uniform sampling is superior only in the special case when clients have the same amount of data.
Fichier principal
Vignette du fichier
A_General_Theory_for_Client_Sampling_in_Federated_Learning.pdf (930.11 Ko) Télécharger le fichier
Origin : Files produced by the author(s)

Dates and versions

hal-03500307 , version 1 (22-12-2021)
hal-03500307 , version 2 (12-07-2022)

Identifiers

  • HAL Id : hal-03500307 , version 2

Cite

Yann Fraboni, Richard Vidal, Laetitia Kameni, Marco Lorenzi. A General Theory for Client Sampling in Federated Learning. International Workshop on Trustworthy Federated Learning in Conjunction with IJCAI 2022 (FL-IJCAI'22), Jul 2022, Vienna, Austria. ⟨hal-03500307v2⟩
193 View
220 Download

Share

Gmail Facebook X LinkedIn More