LKC Trainable linear kernel classifier
W = LKC(A,KERNEL)
DescriptionThis is a fallback routine for other kernel procedures like SVC, RBSVC and LIBSVC. If they fail due to optimisation problems they may fall back to this routine which computes a linear classifier in kernelspace using the pseudoinverse of the kernel. The kernel may be supplied in KERNEL by
If KERNEL = 0 it is assumed that A is already the kernel matrix (square). In this also a kernel matrix should be supplied at evaluation by B*W or PRMAP(B,W). LKC is basically a twoclass classifier. Multiclass problems are solved in a oneagainstrest fashion by MCLASSC. The resulting baseclassifiers are combined by the maximum confidence rule. A better, nonlinear combiner might be QDC, e.g. W = A*(LKC*QDC([],[],1e6)) See alsomappings, datasets, svc, proxm,
