In-network Computation for Large-scale Federated Learning over Wireless Edge Networks

TL;DR

This paper introduces an in-network computation framework for federated learning over wireless edge networks, decentralizing model aggregation to reduce latency and overhead in large-scale FL systems.

Contribution

It proposes a novel in-network aggregation protocol, a joint routing and resource optimization algorithm, and demonstrates significant latency reduction and overhead decrease.

Findings

Training latency reduced up to 5.6 times

Significant decrease in cloud traffic and computing overhead

Effective near-optimal routing algorithm

Abstract

Most conventional Federated Learning (FL) models are using a star network topology where all users aggregate their local models at a single server (e.g., a cloud server). That causes significant overhead in terms of both communications and computing at the server, delaying the training process, especially for large scale FL systems with straggling nodes. This paper proposes a novel edge network architecture that enables decentralizing the model aggregation process at the server, thereby significantly reducing the training delay for the whole FL network. Specifically, we design a highly-effective in-network computation protocol (INC) consisting of a user scheduling mechanism, an in-network aggregation process (INA) which is designed for both primal- and primal-dual methods in distributed machine learning problems, and a network routing algorithm. Under the proposed INA, we then formulate a joint routing and resource optimization problem, aiming to minimize the aggregation latency. The problem is NP-hard, and thus we propose a polynomial time routing algorithm which can achieve near optimal performance with a theoretical bound. Simulation results showed that the proposed INC framework can not only help reduce the FL training latency, up to 5.6 times, but also significantly decrease cloud's traffic and computing overhead. This can enable large-scale FL.