Federated Learning
Also called: distributed learning
Training a shared model across institutions without moving patient data, by exchanging model updates instead of records.
Federated learning sends the model to the data rather than the data to the model. Each site trains locally and shares parameter updates, which are aggregated into a shared model. No patient records leave the institution.
It is attractive in medicine because the main obstacle to large training datasets is legal and institutional rather than technical.
Where This Gets Misread
Federated learning addresses data movement, not data quality or representativeness. It also does not fully solve privacy ... model updates can leak information about training data under some conditions. It is a governance improvement, not a privacy guarantee.