RSSCC

Trainable random subspace combining classifier

     W = RSSCC(A,CLASSF,NFEAT,NCLASSF)
     W = A*RSSCC([],CLASSF,NFEAT,NCLASSF)
     W = A*RSSCC(CLASSF,NFEAT,NCLASSF)

Input
A	Dataset
CLASSF	Untrained base classifier. default NMC
NFEAT	Number of features for training CLASSF
NCLASSF	Number of base classifiers

Output
W	Combined classifer

Description

This procedure computes a combined classifier consisting out of NCLASSF base classifiers, each trained by a random set of NFEAT features of A.  W is just the set of base classifiers and still needs a combiner, e.g.  use W*VOTEC. PRTools uses by default MAXC. A trained combiner might be  considered as well, e.g. W = A*(RSSCC*QDC([],[],1e-6))

The default subspace dimension (NFEAT) is 10% of the size of the training  set A with a minimum of 2. The default number of classifiers is the  dimension of the training set divided by the subspace dimension with a  minimum of 10.

See also

datasets, mappings, parallel, votec, nmc, qdc,