Machine Learning in Elixir

Compressing Data with Autoencoders

If you’re familiar with compression, you already know a bit about unsupervised learning. A somewhat close relationship exists between the objectives of compression and the objectives of unsupervised learning. In a compression algorithm, your goal is to reduce the size of the original data—typically by taking advantage of patterns and structures of the input data. In an unsupervised learning problem, your goal is to capture or model the patterns and structures of the input data.

If you consider a compression algorithm as an Elixir behavior, the contract look something like this:

@callback compress(data) :: compressed_data

@callback decompress(compressed_data) :: data

Recall from Chapter 6, ​Go Deep with ...