Loading…
Glossary term
Training a model across many devices without gathering their raw data in one place.
In federated learning each device or organization trains the model on its own data and sends only the updates, not the data, to a central server. The server averages the updates into an improved shared model.
Raw data never leaves its owner, though updates can still leak information unless extra protections are used.
It enables learning from sensitive or distributed data such as phone keyboards or hospital records.