Deep Learning Book Chapter 14 Notes


What is an autoencoder?

An autoencoder is an approximation of the identity function.

Unlike many approximations, it’s usually meant to be an imperfect one.

An autoencoder is usually the composition of two parts: an encoder and a decoder . So . Generally the encoder is surjective and the decoder is injective, but neither are bijective. In English, the encoder maps the original points into a lower-dimensional space, and the decoder maps them back into a higher-dimensional one. This lower-dimensional bottleneck is where most of the interesting properties of an autoencoder come from.

If you see the standard picture of an autoencoder that makes it look like a tipped-over hourglass, this will make more sense.

Adding a sparsity regularizer such as the norm penalty on the bottleneck layer gives a sparse autoencoder.

Instead of mapping from , we can add some noise to the input and have it try to learn to ignore the noise by giving it the real input as a label.

In math, we use .

The chapter has loads of other stuff, but it didn’t feel interesting enough to me to write down.

Related Posts

How to Disable Disqus Ads on your Blog

Derivation of Reservoir Sampling

Fun with Python Iterators: Linked Lists Made Easy

Notes for November 11, 2018

Underrated Vim Option: undofile and undodir

Hot Take on Solo Travel: Starve

Alan Perlis

Book Notes: The Map of My Life by Goro Shimura

Prague

Way to remember the definition of local finiteness