FedCcs: Federated Learning Cluster-Based Client Selection Algorithm for Non-IID Data
Jingyuan Wang · 2025
Federated learning is a distributed machine learning paradigm that enables collaborative model training across multiple clients without requiring data to leave the local nodes. One of the major challenges in federated learning is data heterogeneity. This paper addresses this issue by proposing a client clustering selection algorithm to mitigate the impact of data heterogeneity. The proposed method is a computationally efficient client selection framework. Our experiments demonstrate that, compared to several existing client selection algorithms, the proposed approach achieves faster convergence and higher test accuracy on public datasets.