Local SGD and Federated Learning: Challenge, Application And Future

DOI

10.2991/978-94-6463-823-3_105How to use a DOI?

Keywords

Local SGD; Federated Learning; Privacy-Preserving Optimization; Non-IID Data

Abstract

Federated learning has emerged as a promising solution to address the critical challenges of privacy protection and communication efficiency in distributed data processing, particularly within the contexts of edge computing and Internet of Things environments. This study explores Local SGD as the primary optimization algorithm for federated learning, providing an in-depth analysis of its theoretical underpinnings, algorithmic advancements, and real-world applications. Through a comprehensive convergence analysis and empirical evaluation of seven optimization variants—including Hybrid Local SGD, SCAFFOLD, and FedProx—we highlight notable progress. Specifically, Hybrid Local SGD achieves 83.76% accuracy on the FEMNIST dataset while reducing communication overhead by 82.36%, and FedProx enhances accuracy by 22% in heterogeneous environments. Practical applications in healthcare (e.g., FedSGDCOVID) and smart energy systems (e.g., FedSign-DP) further validate the effectiveness of Local SGD in striking a balance between privacy, efficiency, and performance. The research also emphasizes Local SGD’s implicit regularization properties and its ability to maintain convergence even in the presence of non-IID data distributions. These findings underscore Local SGD as a foundational framework for privacy-preserving distributed learning, with future research directions focusing on its integration with quantum computing and optimization under more stringent privacy constraints.

Copyright

© 2025 The Author(s)

Open Access

Open Access This chapter is licensed under the terms of the Creative Commons Attribution-NonCommercial 4.0 International License (http://creativecommons.org/licenses/by-nc/4.0/), which permits any noncommercial use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license and indicate if changes were made.

Download article (PDF)

TY  - CONF
AU  - Xiaoxing Tong
PY  - 2025
DA  - 2025/08/31
TI  - Local SGD and Federated Learning: Challenge, Application And Future
BT  - Proceedings of the 2025 3rd International Conference on Image, Algorithms, and Artificial Intelligence (ICIAAI 2025)
PB  - Atlantis Press
SP  - 1090
EP  - 1102
SN  - 2352-538X
UR  - https://doi.org/10.2991/978-94-6463-823-3_105
DO  - 10.2991/978-94-6463-823-3_105
ID  - Tong2025
ER  -