A Genetic Multiple Kernel Relevance Vector Regression Approach

Relevance vector machine (RVM) is a state-of-the-art technique for regression and classification, as a sparse Bayesian extension version of the support vector machine. The kernel function and parameter selection is a key problem in the research of RVM. The real-world application and recent researches have emphasized the requirement to multiple kernel learning. This paper proposes a novel regression technique, called Genetic Multiple Kernel Relevance Vector Regression (GMK RVR), which combines genetic programming and relevance vector regression to evolve a multiple kernel function. The proposed technique are compared with those of a standard RVR using the Polynomial, Gaussian RBF and Sigmoid kernel with various parameter settings, based on several benchmark problems. Numerical experiments show that the GMK performs better than such widely used kernels and prove the validation of the GMK.