Federated Learning via Inexact ADMM

TL;DR

This paper introduces an inexact ADMM algorithm for federated learning that is efficient, robust to stragglers, and converges under mild conditions, outperforming existing methods in numerical tests.

Contribution

It proposes a novel inexact ADMM approach for federated learning that reduces computation and communication costs while ensuring convergence under mild assumptions.

Findings

Outperforms state-of-the-art algorithms in numerical experiments

Capable of handling stragglers effectively

Converges under mild conditions

Abstract

One of the crucial issues in federated learning is how to develop efficient optimization algorithms. Most of the current ones require full device participation and/or impose strong assumptions for convergence. Different from the widely-used gradient descent-based algorithms, in this paper, we develop an inexact alternating direction method of multipliers (ADMM), which is both computation- and communication-efficient, capable of combating the stragglers' effect, and convergent under mild conditions. Furthermore, it has a high numerical performance compared with several state-of-the-art algorithms for federated learning.