PRTools Contents

PRTools User Guide



Weak Classifier


 A Dataset
 ALF Fraction or number of objects to be used for training, see
 GENDAT. Default: one object per class. For integer ALF, ALF objects per class are generated. Default 1;
 N Number of classifiers to be generated, default 1.
 CLASSF untrained classifier, default NMC

 W Best classifier over ITER runs
 V Cell array of all classifiers  Use VC = stacked(V) for combining
 VC Combined set of classifiers


WEAKC uses subsampled versions of A for training. Testing is done  on the entire training set A. The best classifier is returned in W VC combines all classifiers as a stacked combiner in VC.

This routine offers several ways to construct a wea classifier. The  larger ALF, the larger ITER or the more complex CLASSF, the less weak the  resulting classifier W will be. The combined classifier VC is for large  values of N not a weak classifier.

For multi-class problem results may be improved in some cases, e.g.  for small N and / or simple CLASSF, by W = A*(WEKAC*QDC([],[],1e-6))

See also

mappings, datasets, nmc, gendat, stacked,

PRTools Contents

PRTools User Guide

This file has been automatically generated. If badly readable, use the help-command in Matlab.