Proportional Fairness in Federated Learning

TL;DR

This paper introduces a new fairness concept called proportional fairness in federated learning, proposing an algorithm that balances overall and worst-case client performance, validated through extensive experiments.

Contribution

It proposes PropFair, a novel algorithm for achieving proportional fairness in federated learning, with proven convergence and practical effectiveness.

Findings

PropFair approximately finds proportional fairness solutions.

It balances average client performance and worst 10% client performance.

Experimental results validate the effectiveness across vision and language datasets.

Abstract

With the increasingly broad deployment of federated learning (FL) systems in the real world, it is critical but challenging to ensure fairness in FL, i.e. reasonably satisfactory performances for each of the numerous diverse clients. In this work, we introduce and study a new fairness notion in FL, called proportional fairness (PF), which is based on the relative change of each client's performance. From its connection with the bargaining games, we propose PropFair, a novel and easy-to-implement algorithm for finding proportionally fair solutions in FL and study its convergence properties. Through extensive experiments on vision and language datasets, we demonstrate that PropFair can approximately find PF solutions, and it achieves a good balance between the average performances of all clients and of the worst 10% clients. Our code is available at \url{https://github.com/huawei-noah/Federated-Learning/tree/main/FairFL}.