In this article we show how the structure of an autoencoder network can be used to produce the output of an archetypal analysis reducing the computational effort.

PCA improvement and generalization with deep learning:

Dimensionality reduction techniques are common techniques in machine learning and they refer to the process of mapping multidimensional data into a lower dimensional space with minimal loss of information.

The problem of dimensionality reduction is, also, closely related to the problem of features extraction and to the identification of the hidden factors moving the data.

In its basic structure, dimensionality reduction relies on the possibility of mapping data points from a high dimensional input space X to a lower dimensional feature space Y, through a function f: X → Y and back to the original space via a function g: Y → X.

Please complete the following form to access to the free download “Archetypal analysis as an autoencoder”