Trainable automatic radial basis Support Vector Classifier
W = RBSTATSSVC(A)
This routine optimizes a radial basis kernel for STATSSVC. The kernel is based on PROXM. As STATSSVC is basically a two-class classifier, solving multi-class problems in a one-against-rest fashion, different kernels may be needed for every base classifier. The result W is thereby a stacked combiner.
If any class in A has less than 20 objects, the kernel is not optimised but estimated by PKSTATSSVM, using the Parzen kernel.