PKSTATSSVC

Automatic radial basis STATSSVM optimising the Parzen kernel

    [W,KERNEL] = PKSTATSSVC(A,ALF)
    [W,KERNEL] = A*PKSTATSSVC([],ALF)
    [W,KERNEL] = A*PKSTATSSVC(ALF)

Input
A	Dataset
ALF	Parameter, default 0.5

Output
W	Mapping: Radial Basis Support Vector Classifier
KERNEL	Untrained mapping, representing the optimised kernel

Description

This routine provides a radial basis support vector classifier based on  STATSSVC. It uses a radial basis kernel supplied by PROXM with a kernel  width SIGMA found by PARZENC. The kernel width used is ALF*3*SQRT(2)*SIGMA.  This is much faster than the gridsearch used by RBSTATSSVC and performs  about equally well.

Note that STATSSVC is basically a two-class classifier and solves  multi-class problems by MCLASSC, generating a set of one-against-rest  classifiers. PKSTATSSVC, however, supplies a single kernel. For some  problems a proper prescaling of the data, e.g. by SCALEM, amy thereby be  appropriate.

See also

datasets, mappings, statssvc, rbstatssvc, parzenc, svc, mclassc, scalem,