Decentralized Federated Learning on the Edge: From the Perspective of Quantization and Graphical Topology
IEEE Internet of Things Journal · 01.11.2024 · 11 (21) · 34172-34186
DOI: 10.1109/jiot.2024.3400512 ↗Kurz gefasst
Fachartikel in der Zeitschrift IEEE Internet of Things Journal, erschienen am 01.11.2024. Verfasst von Zhigang Yan und 1 weiteren Autor:innen. Die Arbeit wurde bisher 8-mal zitiert (Crossref) und nennt 34 Literaturangaben. Gefördert von: Science and Technology Development Fund, Macau, SAR.
Abstract (Originalsprache)
Decentralized federated learning (DFL), a federated edge learning (FEEL) framework without a server, can avoid the huge communication overhead of the server and the single point of failure within FEEL. Since, there is no server, the convergence of DFL depends not only on the communication conditions and number of edge nodes but also on the graph topology of the network over which edge nodes communicate. Moreover, in practical scenarios, considering the limited communication resources, such as the latency and energy costs, the convergence of DFL also suffers from the quantization errors. In this article, we consider the decentralized gradient descent (DGD), which is widely applied in DFL, and examine the influence of the graph topology and quantization on the convergence of DGD in different wireless networks with the cost constraints. According to the derived convergence bound, the maximum quantization error acceptable for the DFL convergence is obtained. Furthermore, motivated by the limited iterations caused by the impact of constrained communication costs on the each edge node, we compare the convergence bounds between the higher and lower connectivity topologies, which face different energy consumption in one round of communication. Based on this comparison, the impact of the graph topologies with the energy constraints on the convergence can be observed. Numerical simulations confirm the validity of our analyses, supporting the correctness of our theoretical findings.
Förderung
- Science and Technology Development Fund, Macau, SAR
Zitieren
APA
Yan, Z., & Li, D. (2024). Decentralized Federated Learning on the Edge: From the Perspective of Quantization and Graphical Topology. IEEE Internet of Things Journal, 11(21), 34172-34186. https://doi.org/10.1109/jiot.2024.3400512
Vancouver
Yan Z, Li D. Decentralized Federated Learning on the Edge: From the Perspective of Quantization and Graphical Topology. IEEE Internet of Things Journal. 2024;11(21):34172-34186. doi:10.1109/jiot.2024.3400512
BibTeX
@article{Yan2024,
author = {Yan, Zhigang and Li, Dong},
title = {Decentralized Federated Learning on the Edge: From the Perspective of Quantization and Graphical Topology},
journal = {IEEE Internet of Things Journal},
year = {2024},
volume = {11},
number = {21},
pages = {34172--34186},
doi = {10.1109/jiot.2024.3400512}
}
EUVIMED