Abstract: Federated learning enables distributed machine learning while preserving privacy, yet it faces dual challenges of data heterogeneity and device resource constraints. In this paper, we ...