The package provides the python codes to perform a fast forward feature selection using a Gaussian Mixture Model. The algorithm is based on the following papers http://arxiv.org/abs/1501.00857 and Nonlinear parsimonious feature selection for the classification of hyperspectral images. And the code is inspired of https://github.com/mfauvel/FFFS.

Just download the file npfs.py and import it with python. It has been tested on linux, Ubuntu 14.04.

Scipy needs to be installed.

Scikit-learn: >=0.17

For a fast processing, a good linear algebra library is required too. Openblas is a good option.

See example of use in file test.py on artificial data, runTestFull.py on real data when using GMM classifier and runTestSelectionGMM.py on real data when using GMM classifier. (Database is not provided for the last two examples)

Warning: Reference labels need to be integers from 1 to classNumbers.