KT-pFL: A Personalized Federated Learning Model Based on Knowledge Transfer

To address the issues of data privacy protection and the difficulty of federated learning meeting personalized needs in heterogeneous (non-IID) scenarios, this paper proposes a personalized federated learning model based on knowledge transfer, KT-pFL, and completes its experimental reproduction and verification. This model uses the MNIST dataset to construct a strong non-IID experimental scenario. Through knowledge distillation (KD) and personalized parameter retention mechanism, it achieves a balance between global knowledge sharing and local feature adaptation. Comparative experiments with FedAvg show that the local fitting accuracy of KT-pFL reaches 0.9975, and the cross-distribution generalization accuracy is 0.0770. In both performance indicators, it outperforms the benchmark model. The research results validate the effectiveness of the "knowledge distillation + personalized retention" mechanism and provide practical references for federated learning applications in heterogeneous data scenarios.