Skip to content

Autoencoder

Work in progress

These are rough notes, not a finished write-up.

Designed for unsupervised learning, dimensionality reduction, and, in some cases, generative modelling.

Why it is called 'autoencoder'? Trained to encode its input data into a lower-dimensional representation and then decode it back to the original input.

Architecture

Consists of two main parts: an encoder and a decoder.