Spline methods using integration lattices and digital nets (Q843727)

Interpolation and other approximation methods using kernel functions or splines in several dimensions are highly versatile approaches to functional approximation. These approximations are usually associated with a minimisation property which is important to the theoretical analysis, e.g. convergence theorems, and to the practical use. This property also gives rise to a reproducing kernel Hilbert space approach to the analysis. In this paper, such splines are considered and particular choices of the data points are investigated which are useful in applications. These choices include the so-called integration lattices and digital nets and are especially suitable in high dimensions, where normally the approximation problem suffers from the curse of (high) dimensionality. Uniform approximation estimates are derived with the aid of the aforementioned reproducing kernel Hilbert space approach and so-called power functions. Among other set-ups, periodic functions (as approximands) on unit cubes in several dimensions are considered.