FedDA: Resource-adaptive federated learning with dual-alignment aggregation optimization for heterogeneous edge devices

Shaohua Cao, Huixin Wu, Xiwen Wu, Ruhui Ma, Danxin Wang, Zhu Han, Weishan Zhang

Abstract: Highlights•A resource-adaptive federated learning framework for heterogeneous edge devices.•Address both system heterogeneity and data heterogeneity.•Assigning heterogeneous models to clients with different computing capabilities.•Optimizing heterogeneous model aggregation via a dual-alignment optimization method.•Extensive experiments validate the effectiveness of the proposed method.