Feature selection code

Here, you can find implementations (primarily for Matlab/Octave) of feature selection methods appearing in

J. Pohjalainen, O. Räsänen and S. Kadioglu,
"Feature Selection Methods and Their Combinations in High-Dimensional Classification of Speaker Likability, Intelligibility and Personality Traits",
Computer Speech and Language, 29(1), pp. 145-171, 2015 (available online 28 November 2013). pdf

An up-to-date version of the code can be found at our feature selection repository.

Implementations of additional methods appearing in the paper may be included later.

Here is a list of the functions (links point to the repository):
MI.m Mutual information (a feature scoring method)
SD.m Statistical dependency (a feature scoring method)
RSFS.m Random subset feature selection
SFS.m Sequential forward selection
SFFS.m Sequential floating forward selection
KNN.m k-nearest-neighbors classification (for evaluation)
KNNI.m k-nearest-neighbors with incremental distance update (faster for nested feature subsets)
l2i.m A helper function to convert string labels to integer labels
uac.m A scoring function to evaluate the classification performance
demo_simple.m A simple demo script to illustrate the usage of the feature selection functions by using the same data set for feature selection, classifier training and classifier evaluation
demo_cv.m A more realistic demo script using cross validation
ZIP file Contains the current versions

For more information, please see the source code and comments.


Last modified: Thu Mar 26 18:52:05 EET 2015