Positive Semidifinite Kernel Function of SVM Classification Machines under Finite-dimension Space Mapping
有限维空间映射下SVM分类机的半正定核函数

With regard to feature of Support Vector Classification Machines,conditions and results to satisfy mapping of feature space in flnite-dimension are analyzed. The method to determine kernel function satisfied with positive semidefinite condition in mapping space of finite-dimension is put forward. The conclusion is that if Gram matrix which is formed by symmetrical function K is positive semidefinite, the original imported space is certainly mapped into a finitedimension feature space in which the inner product can be expressed by the kernel function K.