A criterion for learning the data-dependent kernel for classification

A novel criterion, namely Maximum Margin Criterion (MMC), is proposed for learning the data-dependent kernel for classification. Different kernels create the different geometrical structures of the data in the feature space, and lead to different class discrimination. Selection of kernel influences greatly the performance of kernel learning. Optimizing kernel is an effective method to improve the classification performance. In this paper, we propose a novel kernel optimization method based on maximum margin criterion, which can solve the problem of Xiong's work [1] that the optimal solution can be solved by iteration update algorithm owing to the singular problem of matrix. Our method can obtain a unique optimal solution by solving an eigenvalue problem, and the performance is enhanced while time consuming is decreased. Experimental results show that the proposed algorithm gives a better performance and a lower time consuming compared with Xiong's work.