A Novel Federated Approach to Model Aggregation for Global Model in Cooperative Wireless Communication

A promising paradigm for decentralized machine learning is federated learning (FL), enabling for local model training yet safeguards data privacy. Despite this, challenges which includes limited resources, varied device capabilities, and communication bottlenecks emerge when integrating FL in wireless communication networks. The hierarchical FL architecture suggested in this research is optimized for multi-hop wireless networks by combining adaptive grouping, dynamic resource allocation, and a two-hop communication protocol to improve accuracy and efficiency. The proposed method’s efficacy in increasing convergence speed, model accuracy, and customer participation rates is supported by experimental findings.