Federated Learning

A machine-learning approach in which a model is trained across multiple devices or servers holding local data, with only model updates, not raw data, shared centrally. Designed to reduce data movement but does not by itself guarantee privacy.