# Variational Autoencoder **Domain:** Generative Modeling / Representation Learning **Doc Type:** Technical Concept Node **Maturity:** Established **Related:** [[Latent Space]], [[Generative Memory Model]], [[Hippocampal Replay]] --- ## Definition A **variational autoencoder (VAE)** is a generative latent-variable model that learns a probabilistic mapping from observations into a latent distribution and a decoder that reconstructs or generates observations from latent samples. ## Memory relevance VAEs provide a formal vocabulary for reconstructive memory models in which experiences are regenerated from latent structure rather than replayed as exact stored copies. A computational analogy does not imply biological identity.