PRTools contents

PRTools manual

minc

MINC

Minimum combining classifier

    W = MINC(V)
    W = V*MINC

Input
 V Set of classifiers

Output
 W Minimum combining classifier on V

Description

If V = [V1,V2,V3, ... ] is a set of classifiers trained on the  same classes and W is the minimum combiner: it selects the class  with the minimum of the outputs of the input classifiers. This  might also be used as A*[V1,V2,V3]*MINC in which A is a dataset to  be classified. Consequently, if S is a dissimilarity matrix with  class feature labels (e.g. S = A*PROXM(A,'d')) then S*MINC*LABELD is the nearest neighbor classifier.

If it is desired to operate on posterior probabilities then the  input classifiers should be extended like V = V*CLASSC;

The base classifiers may be combined in a stacked way (operating  in the same feature space by V = [V1,V2,V3, ... ] or in a parallel  way (operating in different feature spaces) by V = [V1;V2;V3; ... ]

See also

mappings, datasets, votec, maxc, meanc, medianc, prodc, averagec, stacked, parallel,

Example(s)

prex_combining,

PRTools contents

PRTools manual