Support vector machines based on scaling kernels

Types of admissible support vector kernel called scaling kernels are presented in this paper. In fact, scaling kernels are the multi-dimensional scaling function with translation vectors and they are a set of complete bases in the subspace of the square and integrable space. Hence, the goal of support vector machines (SVM) based on scaling kernels is to find the optimal scaling coefficients in a scaling space. In terms of theory, SVM based on scaling kernels can approximate any objective function in some space by any precision. The results obtained by our simulations show the feasibility and validity of scaling kernels.