Energy‐Efficient Federated Learning via Reinforcement Learning‐Based Dynamic Epoch Control
IEEJ Transactions on Electrical and Electronic Engineering · 2026-12 · 21 (12) · 1964-1966
DOI: 10.1002/tee.70324 ↗Kurz gefasst
Fachartikel in der Zeitschrift IEEJ Transactions on Electrical and Electronic Engineering, online erschienen am 28.05.2026. Verfasst von So Tamaki und 1 weiteren Autor:innen. Die Arbeit nennt 5 Literaturangaben. Gefördert von: Japan Society for the Promotion of Science.
Abstract (Originalsprache)
Federated learning enables collaborative model training across distributed clients while preserving data privacy. However, conventional approaches use fixed local epochs on the client side, leading to unnecessary computation when models converge. This study proposes a reinforcement learning (RL)‐based scheme that dynamically controls each client's local epoch count to reduce unnecessary training iterations while maintaining model accuracy. The RL agent learns to avoid assigning redundant epochs as the global model converges, thereby improving training efficiency. Experiments on the Wine Quality dataset with non‐IID data distribution demonstrate that our scheme reduces total local epochs by approximately 60% compared to fixed‐epoch baselines while achieving comparable accuracy, even in the presence of heterogeneous clients. © 2026 Institute of Electrical Engineers of Japan and Wiley Periodicals LLC.
Förderung
- Japan Society for the Promotion of Science
Zitieren
APA
Tamaki, S., & Mizutani, K. (2026). Energy‐Efficient Federated Learning via Reinforcement Learning‐Based Dynamic Epoch Control. IEEJ Transactions on Electrical and Electronic Engineering, 21(12), 1964-1966. https://doi.org/10.1002/tee.70324
Vancouver
Tamaki S, Mizutani K. Energy‐Efficient Federated Learning via Reinforcement Learning‐Based Dynamic Epoch Control. IEEJ Transactions on Electrical and Electronic Engineering. 2026;21(12):1964-1966. doi:10.1002/tee.70324
BibTeX
@article{Tamaki2026,
author = {Tamaki, So and Mizutani, Kimihiro},
title = {Energy‐Efficient Federated Learning via Reinforcement Learning‐Based Dynamic Epoch Control},
journal = {IEEJ Transactions on Electrical and Electronic Engineering},
year = {2026},
volume = {21},
number = {12},
pages = {1964--1966},
doi = {10.1002/tee.70324}
}
EUVIMED