Decentralized and Distributed Learning for AIoT: A Comprehensive Review, Emerging Challenges and Opportunities

Hanyue Xu; Kah Phooi Seng; Li Minn Ang; Jeremy Smith

doi:10.1109/ACCESS.2024.3422211

Back

Decentralized and Distributed Learning for AIoT: A Comprehensive Review, Emerging Challenges and Opportunities

Journal article

Open access

Peer reviewed

Decentralized and Distributed Learning for AIoT: A Comprehensive Review, Emerging Challenges and Opportunities

Hanyue Xu, Kah Phooi Seng, Li Minn Ang and Jeremy Smith

IEEE Access, Vol.12, pp.101016-101052

2024

DOI: https://doi.org/10.1109/ACCESS.2024.3422211

Files and links (1)

pdf

Decentralized_and_Distributed_Learning_for_AIoT_A_Comprehensive_Review_Emerging_Challenges_and_Opportunities5.22 MBDownload View

Published VersionCC BY-NC-ND V4.0, Open Access

Abstract

artificial intelligent Internet of Things

distributed learning

split federated learning

decentralised learning

Artificial intelligence

Graph-based learning

The advent of the Artificial Intelligent Internet of Things (AIoT) has sparked a revolution in the deployment of intelligent systems, driving the need for innovative data processing techniques. Due to escalating data privacy concerns and the immense volume of data produced by IoT devices, decentralized and distributed learning methods that are rapidly replacing traditional centralized learning play a pivotal role. As AIoT systems become increasingly ubiquitous, the accompanying computational and storage demands necessitate a departure from conventional paradigms towards more scalable, distributed, and decentralized architectures. This paper delves into the background of AIoT, with a particular focus on the evolution of distributed and decentralized learning mechanisms that operate without the need for centralized data collection, thus aligning with the General Data Protection Regulation (GDPR) for enhanced data privacy. The various distributed and decentralized learning strategies are the focus of this paper that facilitate collaborative model training across multiple AIoT nodes, thereby not only improving the performance of the AIoT system but also mitigating the risks of data concentration. The review further explores the adaptability of AI algorithms in these distributed settings, assessing their potential to optimize system performance and learning efficacy. The paper concludes with some use cases and lessons learned for decentralized and distributed learning in various AIoT areas.

Details

Title: Decentralized and Distributed Learning for AIoT: A Comprehensive Review, Emerging Challenges and Opportunities
Authors: Hanyue Xu - Xi’an Jiaotong-Liverpool University
Kah Phooi Seng - University of the Sunshine Coast, Queensland, School of Science, Technology and Engineering
Li Minn Ang - University of the Sunshine Coast, Queensland, School of Science, Technology and Engineering
Jeremy Smith - University of Liverpool
Publication details: IEEE Access, Vol.12, pp.101016-101052
Publisher: Institute of Electrical and Electronics Engineers
Date published: 2024
DOI: 10.1109/ACCESS.2024.3422211
ISSN: 2169-3536
Copyright note: This work is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 License. For more information, see https://creativecommons.org/licenses/by-nc-nd/4.
Organisation Unit: School of Science, Technology and Engineering
Language: English
Record Identifier: 991045098802621
Output Type: Journal article

Metrics

3 File views/ downloads

32 Record Views