All terms
Architectures
Autoencoder
A network that compresses input into a small representation and rebuilds the original from it.
Definition
An autoencoder is a neural network trained to squeeze its input into a compact internal representation (the bottleneck) and then reconstruct the original from that code. By forcing the data through a narrow layer, it learns which features matter most. Autoencoders are used for compression, denoising, anomaly detection, and learning embeddings, and the idea underlies parts of modern generative models.