Decentralized Federated Learning over Wireless Channels: A Review

Abstract

Machine learning has made remarkable progress, and a wide range of research has been conducted from theoretical and practical perspectives. However, the expansion of machine learning applications has raised novel concerns, such as data privacy and communication costs. This article outlines distributed machine learning methods such as federated learning (FL), which has been attracting attention in recent years as a method that can both fundamentally solve these problems and analyze big data. Then, we focus on decentralized FL (DFL), which does not have a centralized server. We describe the problems in wireless communication channels and countermeasures to solve them.