Provision for Energy: Federated Learning in Edge Systems

Abstract

An energy-efficient transmission and computation resource allocation problem for federated learning (FL) on wireless communication networks is investigated. Based on the considered model, each user exploits limited local computing resources to train a local FL model with its collected data. The local FL model is then transmitted to a base station (BS), which aggregates the local FL model and broadcasts it back to all users. Based on the learning accuracy level, computation and communication latency are determined by the exchange of learning models between users and BS. During the FL process, both the local computation energy and the transmission energy must be considered. Due to wireless users’ limited energy consumption, the communication problem is formulated as an optimization problem whose objective is to minimize the overall energy consumption of the system with a latency limitation. To solve this problem, we resort to an iterative algorithm with a solution of bandwidth, power, computational and other factors. Numerical results show that the proposed algorithms can reduce energy consumption compared to the conventional FL method.