Federated and Distributed Learning in Edge IoT

Abstract

With the rapid expansion of the Internet of Things (IoT) and the proliferation of smart edge devices, traditional centralized learning models, which collect data into massive data centers for training, become increasingly infeasible due to privacy, latency and bandwidth constraints. Federated Learning (FL) and Distributed Learning (DL) have emerged as promising paradigms that shift intelligence toward the edge, enabling collaborative model training across decentralized devices. This paper reviews the state of the art in federated and distributed learning for edge IoT environments. We explore fundamental concepts, architectural designs, enabling technologies, privacy challenges, communication overhead, and representative applications. We highlight key differences between federated and distributed learning, discuss emerging issues such as model heterogeneity and resource scarcity, and outline future research directions. By synthesizing developments in this field, the paper aims to guide researchers and practitioners seeking scalable, privacy-aware solutions for intelligent edge IoT systems.

KEYWORDS: Federated Learning, Distributed Learning, Edge Intelligence, Internet of Things, Privacy, Edge Computing, Collaborative Learning