# What is MIDAS? Definition and meaning

An imputation method that trains an overcomplete denoising autoencoder and draws multiple completions.

## Definition

**MIDAS**: An imputation method that trains an overcomplete denoising autoencoder and draws multiple completions.

In short: Multiple imputation with denoising autoencoders.

## When it is used

Imputation of mixed numeric and categorical data.

## Learn more

The full guide to **[MIDAS](/tools/midas)** 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: [Microsoft Agent Framework](/glossary/microsoft-agent-framework) · Next term: [Milvus](/glossary/milvus)
