Asynchronous Federated Optimization

TL;DR

This paper introduces an asynchronous federated optimization algorithm that enhances scalability and flexibility, demonstrating near-linear convergence and robustness to staleness across various applications.

Contribution

The paper proposes a novel asynchronous federated optimization algorithm with proven convergence properties and empirical validation, improving scalability and robustness.

Findings

Converges quickly in empirical tests.

Tolerates staleness effectively.

Works for both convex and certain non-convex problems.

Abstract

Federated learning enables training on a massive number of edge devices. To improve flexibility and scalability, we propose a new asynchronous federated optimization algorithm. We prove that the proposed approach has near-linear convergence to a global optimum, for both strongly convex and a restricted family of non-convex problems. Empirical results show that the proposed algorithm converges quickly and tolerates staleness in various applications.