Energy efficiency is essential for federated learning (FL) over mobile devices and its potential prosperous applications. Different from existing communication efficient FL research efforts, which regard communication energy consumption as the bottleneck, we have observed that with ever increasing wireless transmission speed (e.g., Wi-Fi 5 or 5G), the energy consumption of wireless communications for model updates in FL is significantly reduced and sometimes is smaller than that of local on-device training. Motivated by such observations, in this paper, we propose a high-speed wireless communications inspired energy efficient federated learning over mobile devices (EEFL), whose goal is to reduce the overall energy consumption (computing + communication). In particular, we design a novel energy-aware adaptive local update policy for mobile devices by jointly considering FL performance and energy saving of high-speed wireless transmissions. Furthermore, given the device's local update policy in each FL global round, we advance the dynamic voltage and frequency scaling (DVFS) strategy to minimize local training's energy consumption by keeping GPU and CPU working at appropriate frequencies without triggering thermal throttling. Extensive experimental results with various learning models, datasets, and wireless transmission environments demonstrate the proposed EEFL's superiority over the peer designs in terms of energy efficiency.
Bibliographical noteKAUST Repository Item: Exported on 2023-06-19
Acknowledged KAUST grant number(s): BAS/1/1689-01-01, FCC/1/1976-49-01, URF/1/4663-01-01
Acknowledgements: We thank anonymous reviewers and shepherd for their valuable suggestions and feedback. The work of R. Chen and M. Pan was supported in part by NSF under Grant CNS-2107057. The work of Q. Wan and X. Fu was partially supported by NSF under Grants CCF-2130688, CCF-1900904, and CNS-210705. The work of X. Qin was supported by the Young Elite Scientists Sponsorship Program by CAST (Grant No. 2021QNRC001). The work of Y. Hou is partly supported by the National Key R&D Program of China under Grant 2019YFE0114000. The work of D. Wang was supported in part by BAS/1/1689-01-01, URF/1/4663-01-01, FCC/1/1976-49-01 of KAUST, and a funding of the SDAIA-KAUST AI center.