Improving Group Connectivity for Generalization of Federated Deep Learning

arXiv:2402.18949v1 Announce Type: new
Abstract: Federated learning (FL) involves multiple heterogeneous clients collaboratively training a global model via iterative local updates and model fusion. The generalization of FL's global model has a large gap compared with centralized training, which is its bottleneck for broader applications. In this paper, we study and improve FL's generalization through a fundamental ``connectivity'' perspective, which means how the local models are connected in the parameter region and fused into a generalized global model. The term …

abstract applications arxiv connectivity cs.lg deep learning federated learning fusion gap global iterative multiple paper study training type updates via