Heterogeneity-Aware Resource Allocation and Topology Design for Hierarchical Federated Edge Learning

Federated learning (FL) provides a privacy-preserving framework for training machine learning models on mobile-edge devices. Traditional FL algorithms, e.g., FedAvg, impose a heavy communication workload on these devices. To mitigate this issue, hierarchical federated edge learning (HFEL) has been proposed, leveraging edge servers as intermediaries for model aggregation. Despite its effectiveness, HFEL encounters challenges, such as slow convergence rates and high resource consumption, particularly in the presence of system and data heterogeneity. However, existing works are mainly focused on improving the efficiency of traditional FL training, leaving the efficiency of HFEL largely unexplored. In this article, we consider a two-tier HFEL system, where edge devices are connected to edge servers, and edge servers are interconnected through peer-to-peer (P2P) edge backhauls. Our goal is to enhance the training efficiency of the HFEL system through strategic resource allocation and topology design. Specifically, we formulate an optimization problem to minimize total training latency by allocating computation and communication resources, as well as adjusting P2P connections. To ensure convergence under dynamic topologies, we analyze the convergence error bound and introduce a model consensus constraint into the optimization problem. The proposed problem is then decomposed into several subproblems, enabling us to alternatively solve it online. Our method facilitates the efficient implementation of large-scale FL in edge networks under data and system heterogeneity. Experiment results demonstrate that our approach reduces training latency by up to 36.43% under IID data distribution and 35.82% under non-IID distribution, while maintaining model accuracy comparable to baseline methods.