MPPCA

Quick and dirty Python3 implementation of Mixtures of Probabilistic Principal Component Analysers

Description

Mixtures of Probabilistic Principal Component Analysers (MPPCA) is a simple yet powerful algorithm used to cluster data into linear subspaces. Its applications cover clustering, density estimation and classification.

How to use it ?

See the ipython notebook mppca_demo.ipynb.

About this implementation

This implementation is a translation of the matlab implementation Mathieu Andreux and I made in Matlab.

In high dimensions, a logarithmic representation of numbers was required to avoid underflows. Other tricks are well described in the original paper.

Reference

"Mixtures of Probabilistic Principal Component Analysers", Michael E. Tipping and Christopher M. Bishop, Neural Computation 11(2), pp 443–482, MIT Press, 1999

License

WTFPL