A Study of Kernel Discriminant Classify Algorithms Based on CPD Kernels

One of the important techniques is the trick of kernels among the nonlinear classify algorithms.It is a problem which cannot be solved completely till now.For the more rigid restriction conditions of Mercer kernels,substitute them with conditionally positive kernels.The latter has less conditions so as to develop its usage.Some experiments show that c.p.d kernels can be used for several classical nonlinear discriminant classify algorithms such as KPCA and KDDA.As a result,it extends the scope of these classify algorithms.