GPES: An algorithm for evolving hybrid kernel functions of Support Vector Machines

The support vector machine (SVM) is a popular approach to the classification of data. One problem of SVM is how to choose a kernel and the parameters for the kernel. This paper proposes a classification technique, called GPES, that combines genetic programming (GP) and evolutionary strategies (ES) to evolve a hybrid kernel for an SVM classifier. The hybrid kernels are represented as trees that have some adjustable parameters. These hybrid kernels are also the Mercer´s kernels. The experimental results are compared with a standard SVM classifier using the polynomial and radial basis function kernels with various parameter settings.