Personalized Federated Learning with Local Attention. (arXiv:2304.01783v2 [cs.LG] UPDATED)

Federated Learning (FL) aims to learn a single global model that enables the
central server to help the model training in local clients without accessing
their local data. The key challenge of FL is the heterogeneity of local data in
different clients, such as heterogeneous label distribution and feature shift,
which could lead to significant performance degradation of the learned models.
Although many studies have been proposed to address the heterogeneous label
distribution problem, few studies attempt to explore the …

arxiv attention challenge data distribution feature federated learning global learn local attention performance personalized server shift studies the key training