SAFARI: Sparsity-Enabled Federated Learning With Limited and Unreliable Communications
SAFARI: Sparsity-Enabled Federated Learning With Limited and Unreliable Communications
Federated learning (FL) enables edge devices to collaboratively learn a model in a distributed fashion. Many existing researches have focused on improving communication efficiency of high-dimensional models and addressing bias caused by local updates. However, most FL algorithms are either based on reliable communications or assuming fixed and known unreliability …