CLEVALF

Classifier evaluation (feature size curve)

    E = CLEVALF(A,CLASSF,FEATSIZES,TRAINSIZE,NREPS,S,TESTFUN)

Input
A	Training dataset.
CLASSF	The untrained classifier to be tested.
FEATSIZES	Vector of feature sizes (default: all sizes)
TRAINSIZE	Number of objects/fraction of training set size (see GENDAT) or generator mapping.
NREPS	Number of repetitions (default: 1)
S	Independent test dataset (optional)
TESTFUN	Mapping,evaluation function (default classification error)

Output
E	Structure with results  See PLOTE for a description

Description

Generates at random for all feature sizes stored in FEATSIZES training  sets of the given TRAINSIZE out of the dataset A. See GENDAT for the  interpretation of TRAINSIZE. These are used for training the untrained  classifier CLASSF. The result is tested by all unused ojects of A, or,  if given, by the test dataset S. This is repeated N times. If no testset  is given and if LEARNSIZE is not given or empty, the training set is  bootstrapped. If a testset is given, the default training set size is  the entire training set. Default FEATSIZES: all feature sizes.  The mean erors are stored in E. The observed standard deviations are  stored in S. The default test routine is classification error estimation  by TESTC([],'crisp').

See CLEVALFS for how to construct feature curves in addition with  automatic feature extraction / selection.

This function uses the RAND random generator and thereby reproduces only  if its seed is saved and reset.

See also

mappings, datasets, cleval, clevalfs, testc, plote, gendat,