Federated Learning with Support of HetNets, Cloud Computing, and Edge Computing


Koçyiğit A. , Ever E.

in: Real-Time Intelligence for Heterogeneous Networks, Fadi Al-Turjman, Editor, Springer Nature, Zug, pp.19-42, 2021

  • Publication Type: Book Chapter / Chapter Research Book
  • Publication Date: 2021
  • Publisher: Springer Nature
  • City: Zug
  • Page Numbers: pp.19-42
  • Editors: Fadi Al-Turjman, Editor

Abstract

With the recent advancements in heterogeneous networks, particularly following the improvements in the Internet of Things (IoT) supporting infrastructures, various machine learning applications which use distributed computing facilities such as cloud, fog, and edge computing have gained popularity. One way of performing computationally intensive learning-related tasks is through distributed machine learning. Due to certain privacy-related concerns, it may not be possible to collect data representative enough to fit a generalisable machine learning model. In such cases, decentralised approaches such as federated learning become a viable option. Federated learning techniques can be used effectively by employing large numbers of participants of heterogeneous nature in terms of computational and storage resources, communication interfaces, as well as types and volumes of available data. In this chapter, federated learning and related concepts are explored together with heterogeneous networks and heterogeneous objects. Enabling technologies such as cloud, fog, edge, and mobile edge computing facilities are discussed together with federated learning architectures. Existing challenges and open research issues are considered critically, taking the heterogeneous nature of federated learning into account, particularly for the applications in the field of IoT.