AI Glossary

Latent space

Also called latent representation

What does “Latent space” mean?

Latent space: The compressed, internal set of numbers a model uses to represent data. Points that are close together in this space tend to share meaning or features, even though the individual dimensions rarely correspond to anything humans can name directly.

Latent space visual guide: The compressed, internal set of numbers a model uses to represent data. Points that are close together in this space tend to share meaning or features, even though the individual dimensions rarely correspond to anything humans can name directly.
Visual guide Open full size (new tab)

What is an example of “Latent space”?

Image generators often work in latent space instead of on raw pixels, which is much cheaper, and moving smoothly between two points can blend one image into another.

Keep learning

See also

EmbeddingsTermVariational autoencoderTermLatent diffusionTermEmbedding modelTerm

Same topic

Generative adversarial networkTermLong short-term memoryTermGPTTermMixture of expertsTermMulti-head attentionTerm