Tensor-product approximation to operators and functions in high dimensions (Q2465299)

The paper concerns an efficient approximation of multi-variate functions and function-related tensors and certain multidimensional integral operators. The authors analyse collocation-type schemes for separable approximation of functions and tensors, which are based on tensor product functions. Constructive schemes are described via the sinc-quadrature and the sinc-interpolation methods. Questions of error bounds and numerical complexity of the methods are considered. The results are applied to the structured tensor-product representation of integral operators with some shift-invariant kernels in many spatial domains. As an example, the classical Newton potential is considered. Numerical illustration are given for the dimension \(d=3\). The paper is a continuation of the authors' recent papers [Computing 76, No.~3--4, 203--225 (2006; Zbl 1087.65050); J. Numer. Math. 13, No.~2, 119--156 (2005; Zbl 1081.65035); Comput. Methods Appl. Math. 6, No.~2, 194--220 (2006; Zbl 1120.65052)].