Deep Learning

Latent Space, explained

Latent space is the internal representation space where a model encodes hidden factors or learned structure.
Back to Glossary

Part of the Semantic Notion technical glossary.

Short Definition

Latent space is the internal representation space where a model encodes hidden factors or learned structure.

Intuition

It is the model's compressed map of what matters about the input, even if those dimensions are not directly human-labeled.

Technical Definition

A latent space consists of learned variables or vectors that represent unobserved factors inferred from data by a model.

Example

An autoencoder may place visually similar images near each other in latent space.

Common Misunderstandings

Latent dimensions do not always have clean human meanings.
Distance in latent space is only meaningful relative to the training objective.

Start here

Need the broader concept map?

Return to the glossary index to move from this definition into adjacent AI, ML, mathematics, and computer science terms.