Fast evaluation of radial basis functions: Methods for generalized multiquadrics in \(\mathbb R^n\) (Q2780593)

Radial basis functions are a useful tool for functional approximation in several dimensions when scattered data are provided. Their convergence properties are very good, but the implementation of radial basis functions methods is difficult. One of the main problems, namely the need for fast evaluation methods, is overcome by far field expansion schemes such as those studied by the authors in this paper. NEWLINENEWLINENEWLINEThe novelties lie in the applicability to multiquadric radial basis functions (and their generalisations) and in more than two dimensions. Almost optimal schemes are proposed that admit the fast evaluation and computation of interpolants to scattered data by hierarchical fast evaluators in any dimension. Concrete algorithms are provided and tests are recorded as well.