Abstract: The kernel optimal transformation and cluster centers algorithm (KOT-CC) is presented by using kernel methods. In the KOT-CC, all data are mapped to a kernel space via some nonlinear mapping and the optimal transformation and cluster centers (OT-CC) is performed in the kernel space. KOT-CC is a powerful technique for extracting nonlinear discriminant features and is very effective in solving pattern recognition problems which have serious overlap between the patterns of different classes. A fast algorithm for KOT-CC is also proposed based on the basis of the sub-space which is spanned by the training samples mapped into the kernel space, which can improve the efficiency of the feature extraction process and tackle the “large sample size” problem which many kernel methods may suffer from. The experiments based on the data of IRIS, YEAST, GLASS and so on, demonstrate the validity of the proposed new algorithm.

Key words: kernel methods, optimal cluster centers, optimal transformation