I would like to compare methods for selecting ridge parameter in kernel regression. To do that, I want to simulate Kernel Gram matrices with different kernels (polynomials, Gaussian.) which satisfy a fast decreasing of the eigenvalues.