Approximating Mutual Information for Multi-label Feature Selection

Abstract [Title Page]

This work focuses on approximating mutual information for multi-label feature selection. The proposed method selects a feature subset that maximizes dependency to target labels and minimizes inter-dependency among features.

The main technical ideas behind how this program works appear in this paper:

Jaesung Lee, Hyunki Lim, and Dae-Won Kim, "Approximating Mutual Information for Multi-label Feature Selection," Electronics Letters, 2012

This software is a Matlab implementation of the proposed method, specialized on problems of multi-label feature selection. The original version of this program was written by Jaesung Lee and Hyunki Lim.


Bibtex Code

@article{lee2012approximating,
  title={Approximating mutual information for multi-label feature selection},
  author={Lee, Jaesung and Lim, Hyunki and Kim, Dae-Won},
  journal={Electronics letters},
  volume={48},
  number={15},
  pages={929--930},
  year={2012},
  publisher={IET}
}


Download

This program is available for download for non-commercial use, licensed under the GNU General Public License, which is allows its use for research purposes or other free software projects but does not allow its incorporation into any type of commerical software.

Download (2012-07-12)

The zipped package includes componenet for source file.


Sample Input and Output

It will find the index of input categorical features in orders of proposed method. This code can executed under Matlab command window.

[Usage]:
   >> index = proposed( data, label, number );

[Description]
   data - a matrix that is composed of features
   label - a matrix that is composed of labels(multi-label)
   number - the number of features to be selected
   index - an index of selected features

By convention in the input features matrix, rows represent data(patterns) and columns represent features.