KPCA Procedure

reads input data in parallel when the data source is on a distributed system

is multithreaded during all phases of analytic execution

supports large-scale training data

enables you to choose among linear, polynomial, and radial basis function (Gaussian) kernels

enables you to perform fast training and fast scoring of KPCA models

enables you to perform the analysis on selected columns of the data

enables you to project new data onto principal components that have captured the nonlinear relationship in the data

enables you to compute the pre-image of the points in the kernel principal component space back into the original input space