# What is Federated learning? Definition and meaning

Each site trains locally and only model updates are aggregated centrally, so raw records never leave their owner.

## Definition

**Federated learning**: Each site trains locally and only model updates are aggregated centrally, so raw records never leave their owner.

In short: Training across sites without moving the raw data.

## When it is used

Privacy-constrained settings such as hospitals, banks or factories that cannot pool data.

## Learn more

The full guide to **[Federated learning](/tools/federated-learning)** explains how it works, when to use it and how it compares with the alternatives, with examples. The [glossary](/glossary) lists every term on the site from A to Z.

Previous term: [Feature store](/glossary/feature-store) · Next term: [Few-shot learning](/glossary/few-shot-learning)
