Deep learning
Autoencoder
An autoencoder learns to reconstruct an input through a constrained latent representation.
How it works
The encoder compresses the input and the decoder rebuilds it. The constraint encourages the latent code to retain useful structure rather than copy every value directly.
A concrete example
Illustrative: compress 784 image pixels into 32 latent values, then reconstruct 784 pixels.
Do not confuse it with Variational autoencoder
a basic autoencoder maps to a point; a VAE learns a latent distribution.