Asynchronous SGD on Graphs: a Unified Framework for Asynchronous Decentralized and Federated Optimization

TL;DR

This paper introduces AGRAF SGD, a unified asynchronous decentralized optimization framework that combines asynchronous communication and decentralization, improving convergence rates under milder assumptions for various algorithms.

Contribution

The paper presents AGRAF SGD, a general framework unifying asynchronous decentralized and federated optimization algorithms with improved convergence guarantees.

Findings

Provides convergence rates under milder assumptions.

Recovers or improves upon best known results for multiple algorithms.

Unifies various asynchronous decentralized algorithms under one framework.

Abstract

Decentralized and asynchronous communications are two popular techniques to speedup communication complexity of distributed machine learning, by respectively removing the dependency over a central orchestrator and the need for synchronization. Yet, combining these two techniques together still remains a challenge. In this paper, we take a step in this direction and introduce Asynchronous SGD on Graphs (AGRAF SGD) -- a general algorithmic framework that covers asynchronous versions of many popular algorithms including SGD, Decentralized SGD, Local SGD, FedBuff, thanks to its relaxed communication and computation assumptions. We provide rates of convergence under much milder assumptions than previous decentralized asynchronous works, while still recovering or even improving over the best know results for all the algorithms covered.