Abstract: Decentralized federated learning (DFL) is a promising technique to enable distributed machine learning over edge nodes without relying on a centralized parameter server. However, existing ...