Radial basis function approximation of noisy scattered data on the sphere (Q1679208): Difference between revisions

Under the choice of a variety of computations for smoothing parameters, a least-squares smoothing (often used in practice in connection with cross-validation) is computed to noisy data using radial basis functions. The radial basis function approach to approximation is an approach of much success and is here used on spheres. The particular advantage of radial basis function approximation is, among others, that interpolants always exist under mild conditions on the data points (e.g.\ that they be distinct) and that the approximants from the spaces spanned by them minimize naturally a so-called native space (\(L^2\)) norm. For the purpose in the present article, geodetic distances replace Euclidean distances used otherwise in \(R^n\). The parameters for the radial basis function smoothing are chosen by one a posteriori and three a priori schemes; numerical examples are included.