Personalized Federated Learning with Mixture of Models for Adaptive Prediction and Model Fine-Tuning

Most previous work on federated learning assumes that clients possess static batches of training data. However, clients may also need to make real-time predictions on streaming data in non-stationary environments.