Kernel \(B\)-splines and interpolation (Q2490413)

Kernel B-splines comprise a more general approach to radial basis function interpolation, in that the more specific radial basis functions are replaced by positive definite kernels or conditionally positive kernels. In this paper, differences are applied to the conditionally positive definite kernels in order that they decay (more quickly) and that the linear systems for interpolation are easier to solve and solve more stably. General data sets of ``centres'' are allowed and also adaptive methods to construct the very useful kernels are derived.