MODSELC

Model selection

    V = MODSELC(A,W,N,NREP)
    V = A*(W*MODSELC([],N,NREP))

Input
A	Dataset used for training base classifiers and/or selection
B	Dataset used for testing (executing) the selector
W	Set of trained or untrained base classifiers
N	Number of crossvalidations, default 10
NREP	Number of crossvalidation repetitions

Output
V	Selected classidfer

Description

This routine selects out of a set of given classifiers stored in W the  best one on the basis of N-fold crossvalidation (see CROSSVAL), which  might be repeated NREP times. If W contains a set of already trained  classifiers, N and NREP are neglected and just the best classifier  according to the evaluation set A is returned.

This routine can be considered as a classifier combiner based on global  selection. See DCSC for local, dynamic selection.

See also

datasets, mappings, stacked, dcsc, classc, testd, labeld,