Energy-Efficient Online Node Cooperation Strategy for Hierarchical Federated Learning
Sailan Zou, Zhuo Li, Xin Chen · 2022
With the popularity of Internet of Things (IoT), a large amount of data is generated at the network edge. Federated Learning (FL) can use this data for joint Cloud-Client model training. The overhead of uploading data in FL is high, and using edge servers for partial model aggregation in Hierarchical Federated Learning (HFL) can reduce the latency and energy cost. In HFL, opportunistic communication and D2D provide opportunities for node cooperation. In this work, we optimize the node cooperation strategy using opportunistic communication with the objective to minimize energy cost under the delay constraint. We design an online node cooperation strategy (OSRN) based on the optimal stopping theory. Through theoretical analysis, we prove the NP-hardness of the problem investigated. We conduct through simulation experiments and find that the proposed algorithm outperforms the random selection algorithm with 18.09% reduction in energy cost.