On shifted cardinal interpolation by Gaussians and multiquadrics (Q2564365): Difference between revisions

The authors study so-called shifted Gaussian interpolation where approximants are given as a (finite or infinite) linear combination of translates of some Gaussian. More precisely, the approximants \(s\) are of the form \[ s(x)= \sum_{k\in\mathbb{Z}^d} a_k \varphi(x+\alpha-k) \qquad (x\in\mathbb{R}^d), \tag{1} \] where \(\mathbb{Z}\) is the set of integers, the (shift) \(\alpha\) is a fixed vector in \(\mathbb{R}^d\) and \(\varphi(x)= \varphi_\lambda(x)= \exp(-\lambda|x|^2_2)\) with a positive parameter \(\lambda\) is a Gaussian. Of particular interest are the cardinal functions \(\chi_\alpha\) being of the form (1) for certain \(a_k(\alpha)\) and satisfying \(\chi_\alpha(j)= \delta_{0j}\) for \(j\in \mathbb{Z}^d\). One main result is that such \(\chi_\alpha\) exist if \(\alpha\) has no half-integer components. The proof is founded on the analysis of the Toeplitz operator \((\varphi(\alpha+j- k))_{j,k\in\mathbb{Z}^d}\) using its links to the theory of Jacobian theta functions. Moreover, it is shown that for such \(\alpha\) the shifted cardinal interpolants to a square-integrable function \(f\) converge to this function (for \(\lambda\) tending to 0) in the mean-square sense if and only if \(f\) is band limited.