One-Shot Federated Learning
This addresses the challenge of communication efficiency in federated learning for distributed networks, though it appears incremental as it builds on ensemble learning and knowledge aggregation.
The paper tackled the problem of training a global model across federated devices with minimal communication, achieving a 51.5% average relative gain in AUC over local baselines and reaching 90.1% of the ideal global performance in a single communication round.
We present one-shot federated learning, where a central server learns a global model over a network of federated devices in a single round of communication. Our approach - drawing on ensemble learning and knowledge aggregation - achieves an average relative gain of 51.5% in AUC over local baselines and comes within 90.1% of the (unattainable) global ideal. We discuss these methods and identify several promising directions of future work.