Federated Learning With Adjustable Learning Rates for Resource-Constrained Wireless Networks

Wireless federated learning (WFL) suffers from heterogeneity prevailing in data distributions, computing powers, and channel conditions of participating devices. This paper presents a new Federated Learning with Adjusted leaRning ratE (FLARE) framework to mitigate the impact of the heterogeneity. The key idea is to allow the participating devices to adjust their individual learning rates and local training iterations, adapting to their instantaneous computing powers. The convergence upper bound of FLARE is established rigorously under a generic setting with non-convex models, non-i.i.d. datasets and imbalanced computing powers. By minimizing the upper bound, we further optimize the scheduling strategy of FLARE to exploit the channel heterogeneity. A nested problem structure is uncovered to facilitate iterative bandwidth allocation with binary search and device selection with a new greedy method. Experiments demonstrate that FLARE consistently outperforms the baselines in test accuracy, and converges much faster with the proposed scheduling policy.