A set of new kernel function for support vector machines: An approach based on Chebyshev polynomials

In this paper, we introduce a set of new kernel functions Which is derived by combining generalized Chebyshev polynomials with other standard kernel functions. New kernel functions have significant advantages over classic support Vector Machine's (SVM) kernel functions and Chebyshev kernel. Simulation results illustrate the fact that the new set of kernel functions (in particular Chebyshev-Gaussian kernel) has noticeable improvement in decreasing error rate and support vector numbers.