Federated Learning Framework via Distributed Mutual Learning
This work addresses privacy and efficiency issues in Federated Learning for distributed machine learning applications, though it appears incremental as it builds on existing mutual learning concepts.
The paper tackled the problem of network bandwidth burden and privacy risks in Federated Learning by proposing a loss-based alternative using distributed mutual learning, which achieved higher accuracy on unseen data with stronger generalization and privacy benefits in face mask detection experiments.
Federated Learning often relies on sharing full or partial model weights, which can burden network bandwidth and raise privacy risks. We present a loss-based alternative using distributed mutual learning. Instead of transmitting weights, clients periodically share their loss predictions on a public test set. Each client then refines its model by combining its local loss with the average Kullback-Leibler divergence over losses from other clients. This collaborative approach both reduces transmission overhead and preserves data privacy. Experiments on a face mask detection task demonstrate that our method outperforms weight-sharing baselines, achieving higher accuracy on unseen data while providing stronger generalization and privacy benefits.