A Study of Enhancing Federated Learning on Non-IID Data with Server Learning
Authors: Van Sy Mai, Richard J. La, Tao Zhang
National Institute of Standards and Technology (NIST) · University of Maryland
To appear in IEEE Transactions on Artificial Intelligence
Shows that auxiliary server learning, using a small server-side dataset, improves federated learning accuracy and convergence when client data is non-IID, and also helps SCAFFOLD, FedDyn and FedDC.
Original work by Van Sy Mai, Richard J. La, Tao Zhang. Showcased by ArtificialIntelligence.com.ai, not authored by us.