feature subset selection and feature ranking
[idx,score] = nan_fss(D,cl) [idx,score] = nan_fss(D,cl,MODE) [idx,score] = nan_fss(D,cl,MODE)
data - each column represents a feature
classlabel
[default] correlation
correlation
feature selection algorithm based on a distance discriminant [2]
max-relevance, min redundancy [1] - not supported yet.
score of the feature
ranking of the feature
the method is motivated by the max-relevance-min-redundancy (mRMR) approach [1]. However, the default method uses partial correlation, which has been developed from scratch. PCCM [3] describes a similar idea, but is more complicated. An alternative method based on FSDD is implemented, too.
[1] Peng, H.C., Long, F., and Ding, C.,
Feature selection based on mutual information: criteria of max-dependency, max-relevance, and min-redundancy,
IEEE Transactions on Pattern Analysis and Machine Intelligence,
Vol. 27, No. 8, pp.1226-1238, 2005.
[2] Jianning Liang, Su Yang, Adam Winstanley,
Invariant optimal feature selection: A distance discriminant and feature ranking based solution,
Pattern Recognition, Volume 41, Issue 5, May 2008, Pages 1429-1439.
ISSN 0031-3203, DOI: 10.1016/j.patcog.2007.10.018.
[3] K. Raghuraj Rao and S. Lakshminarayanan
Partial correlation based variable selection approach for multivariate data classification methods
Chemometrics and Intelligent Laboratory Systems
Volume 86, Issue 1, 15 March 2007, Pages 68-81
http://dx.doi.org/10.1016/j.chemolab.2006.08.007