Source: wikibot/variational-autoencoder

= Variational autoencoder
{wiki=Variational_autoencoder}

A Variational Autoencoder (VAE) is a type of generative model that is used in unsupervised machine learning tasks to learn the underlying structure of data. It combines principles from probabilistic graphical models and neural networks. Here are the key components and ideas behind VAEs: \#\#\# Structure A VAE typically consists of two main components: 1. **Encoder (Recognition Model)**: This part of the VAE takes input data and encodes it into a lower-dimensional latent space.