Method of good matrices for multi-dimensional numerical integrations - An extension of the method of good lattice points (Q1099925): Difference between revisions

A method of ``good matrices'' is developed for numerical integration over the whole of \({\mathbb{R}}^ s.\) The rule samples the integrand with equal weights at points obtained from mapping the integer points by a matrix A. The method may be regarded as an extension of \textit{K. K. Frolov}'s method [Dokl. Akad. Nauk SSSR 232, 40-43 (1977; Zbl 0368.65016)] which itself is a generalization of the familiar method of good lattice points. The Poisson summation formula is used to obtain error estimates for classes of functions whose Fourier transforms decay in prescribed ways at infinity. Matrices that optimize or nearly optimize dependence on A are ``good matrices''. Known results from the geometry of numbers assist with the choices. Some numerical experiments for \(s\leq 4\) demonstrate the usefulness of the method.