AFA-Fed: Adaptive Fast Asynchronous Federated Learning for Resource-Constrained Edge Devices
IEEE Transactions on Consumer Electronics · 2026-05 · 72 (2) · 3123-3135
DOI: 10.1109/tce.2026.3682624 ↗Kurz gefasst
Fachartikel in der Zeitschrift IEEE Transactions on Consumer Electronics, 2026-05. Verfasst von Li Zhang und 5 weiteren Autor:innen. Die Arbeit nennt 34 Literaturangaben. Gefördert von: Natural Science Foundation of Hunan Province, National Undergraduate Training Program for Innovation and Entrepreneurship, Guangdong Province Key Discipline Scientific Research Capability Improvement Project.
Abstract (Originalsprache)
The essence of federated learning lies in collaboratively training a model across multiple clients while preserving privacy. Implementing efficient federated learning in edge computing faces various challenges, including device heterogeneity, resource constraints, and dynamic network environments in consumer electronics (e.g., smart home systems and wearables). This paper proposes an adaptive federated optimization algorithm (AFA-Fed) based on Deep Deterministic Policy Gradient (DDPG) from reinforcement learning. The algorithm aims to minimize global training time while considering resource constraints and model convergence guarantees. It intelligently adjusts the number of local models participating in global model aggregation in each round based on different federated learning tasks and real-time network conditions, guiding the central server to optimize the selection of training nodes and resource allocation strategies. Additionally, a fast edge device selection strategy is introduced to mitigate the impact of slow clients. Finally, experiments on two real-world datasets validate the effectiveness and reliability of the proposed solution. Under resource-constrained consumer settings, AFA-Fed can reduce training time by approximately 45% and improve training accuracy by about 25% compared to state-of-the-art methods.
Förderung
- Natural Science Foundation of Hunan Province
- National Undergraduate Training Program for Innovation and Entrepreneurship
- Guangdong Province Key Discipline Scientific Research Capability Improvement Project
- Innovation Project of Universities in Guangdong Province
- Doctoral Research Startup Fund of Hunan Institute of Engineering
Zitieren
APA
Zhang, L., Xiao, Y., Tang, B., He, W., Huang, Y., & Li, K. (2026). AFA-Fed: Adaptive Fast Asynchronous Federated Learning for Resource-Constrained Edge Devices. IEEE Transactions on Consumer Electronics, 72(2), 3123-3135. https://doi.org/10.1109/tce.2026.3682624
Vancouver
Zhang L, Xiao Y, Tang B, He W, Huang Y, Li K. AFA-Fed: Adaptive Fast Asynchronous Federated Learning for Resource-Constrained Edge Devices. IEEE Transactions on Consumer Electronics. 2026;72(2):3123-3135. doi:10.1109/tce.2026.3682624
BibTeX
@article{Zhang2026,
author = {Zhang, Li and Xiao, Yuqiang and Tang, Bing and He, Wenhui and Huang, Yuan and Li, Kuanching},
title = {AFA-Fed: Adaptive Fast Asynchronous Federated Learning for Resource-Constrained Edge Devices},
journal = {IEEE Transactions on Consumer Electronics},
year = {2026},
volume = {72},
number = {2},
pages = {3123--3135},
doi = {10.1109/tce.2026.3682624}
}
EUVIMED