TESTD

Find classification error of dataset applied to given classifier

    [E,C] = TESTD(A,W)
    [E,C] = TESTD(A*W)
    [E,C] = TESTD(A,LABELS)
     E = A*W*TESTD

Input
A	Dataset used for testing
W	Classifier
LAB	Estimated labels

Output
E	Fraction of labeled objects in A that is erroneously classified or  that have different labels than LAB.
C	Vector with numbers of erroneously classified objects per class.  They are sorted according to A.LABLIST

Description

This routine performs a straightforward testing of the labeled objects  in A by the classifier W. Unlabeled objects are neglected. Just a  fraction of the erroneously classified objects is returned. Not all  classes have to be present in A and class priors are neglected. TESTD thereby essentially differs from TESTC which tests the classifier instead  of the dataset. Labels found for the individual objects by the classifier  W may be retrieved by LABELD.

The dataset A should have crisp labels. If needed they are converted to  crisp first.

See also

datasets, mappings, testc, labeld,