# What is GAT? Definition and meaning

A graph network that learns attention weights over each node's neighbours.

## Definition

**GAT**: A graph network that learns attention weights over each node's neighbours.

In short: Graph attention network.

## When it is used

Graph tasks where neighbours differ in importance.

## Learn more

The full guide to **[GAT](/tools/gat)** 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: [GAN](/glossary/gan) · Next term: [Gaussian splatting](/glossary/gaussian-splatting)
