Choosing shape parameters for regression in reproducing kernel Hilbert space and variable selection

The Gaussian radial basis function (RBF) is a widely used kernel in kernel-based methods. parameter RBF, referred to as the shape parameter, plays an essential role model fitting. In this paper, we propose method select parameters for general RBF kernel. It can simultaneously serve variable selection and regression estimation. For former, asymptotic consistency established; latter, estimation efficient if true or optimal are known. Simulations real examples illustrate method's performance of prediction by comparing it with other popular