Skip to content

Autoencoders

Autoencoders (AE) are machines that encode inputs into a compact latent space.

AE

Notation: dot ()

We use a single vertically centered dot, i.e., , to indicate that the function or machine can take in arguments.

A simple autoencoder can be achieved using two neural nets, e.g.,

(1)h=g(b+wx)(2)x^=σ(c+vh),

where in this simple example,

  • g(b+w) is the encoder, and
  • σ(c+v) is the decoder.

For binary labels, we can use a simple cross entropy as the loss.

Code

See Lippe1.


  1. Lippe P. Tutorial 9: Deep Autoencoders — UvA DL Notebooks v1.1 documentation. In: UvA Deep Learning Tutorials [Internet]. [cited 20 Sep 2021]. Available: https://uvadlc-notebooks.readthedocs.io/en/latest/tutorial_notebooks/tutorial9/AE_CIFAR10.html 


Contributors: LM