Decentralized Personalized Federated Learning based on a Conditional Sparse-to-Sparser Scheme

arXiv:2404.15943v1 Announce Type: new
Abstract: Decentralized Federated Learning (DFL) has become popular due to its robustness and avoidance of centralized coordination. In this paradigm, clients actively engage in training by exchanging models with their networked neighbors. However, DFL introduces increased costs in terms of training and communication. Existing methods focus on minimizing communication often overlooking training efficiency and data heterogeneity. To address this gap, we propose a novel \textit{sparse-to-sparser} training scheme: DA-DPFL. DA-DPFL initializes with a subset of model parameters, …

abstract arxiv become communication costs cs.ai cs.lg decentralized federated learning focus however neighbors paradigm personalized popular robustness terms training type