Autoencoder

A type of artificial neural network used to learn efficient codings of unlabeled data (unsupervised learning).

How it works

An autoencoder has two parts: an encoder that compresses the input into a lower-dimensional latent representation, and a decoder that reconstructs the original input from that compressed form. The model is trained to minimise the difference between the input and the reconstruction. Because it must squeeze information through a narrow bottleneck, the encoder is forced to learn only the most essential features of the data.

Why it matters

Autoencoders underpin a wide range of applications including anomaly detection (unusual inputs that compress poorly), data denoising, and dimensionality reduction. They are also the architectural ancestor of the Variational Autoencoder (VAE), which is a key component of many image generation systems. Understanding autoencoders gives insight into how modern generative models learn compact, meaningful representations of complex data.