Improving Federated Learning Performance by Leveraging Client Data Diversity

Clients with more diverse data can improve the performance of federated learning models. By emphasizing updates from high-diversity clients and diminishing the influence of low-diversity clients, the proposed WeiAvgCS framework can significantly improve the convergence speed of federated learning.