关键词: 联邦学习, 原型聚类, 费舍尔信息矩阵, 加权融合, 客户端选择

Abstract: Federated learning, as an important research direction in the field of artificial intelligence, was utilized to construct global models through distributed collaborative training under the premise of ensuring data privacy, providing a new paradigm for joint modeling in privacy-sensitive scenarios. However, existing methods were challenged by performance degradation, training oscillation, and slow convergence when faced with heterogeneous client data. To address these issues, a federated learning algorithm based on prototype clustering and adaptive weighted fusion, named FedPFA, was proposed. First, the prototypes uploaded by clients were clustered by the server to mitigate data distribution discrepancies and enhance the consistency and generalization capability of the global model. Subsequently, differentiated weights were assigned to the prototypes by incorporating the trace estimates of the Fisher information matrix from each client, thereby highlighting the contributions of high-quality clients and suppressing interference from noisy or inadequately trained clients during the fusion process. Experimental results on the MNIST, Fashion-MNIST, and CIFAR-10 datasets demonstrated that the accuracy was improved by approximately 19.64, 26.14, and 16.15 percentage points, respectively, compared to the FedGH method. The proposed algorithm not only ensured performance enhancement but also significantly improved the convergence speed and stability of the global model, exhibiting strong robustness and practical application value across various datasets.

Key words: federated learning, prototype clustering, fisher information matrix, weighted fusion, client selection