Introduction
Your phone's keyboard gets better at predicting your next word. But the companies building these keyboards don't want to — and often legally can't — upload everything you type to their servers.
Federated learning (FL) solves this: instead of bringing the data to the model, it brings the model to the data. Each device (or hospital, or bank branch) trains the model locally on its own data and sends back only model updates — never the raw data. A central server combines the updates into an improved shared model.
Where it's used:
- mobile keyboards — next-word prediction and emoji suggestions;
- healthcare — several hospitals train a diagnosis model together without sharing patient records;
- finance — banks collaborate on fraud detection;
- IoT and cars — learning from many devices at the edge.