# What is MAE? Definition and meaning

Masks a large part of the input and trains the model to reconstruct it, learning rich representations.

## Definition

**MAE**: Masks a large part of the input and trains the model to reconstruct it, learning rich representations.

In short: Masked autoencoder: reconstruct hidden patches.

## When it is used

The vision counterpart of masked modelling, and the intuition behind masked cell modelling for tables.

## Learn more

The full guide to **[MAE](/tools/mae)** 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: [LwF](/glossary/lwf) · Next term: [Mamba](/glossary/mamba)
