Optimal complexity recovery of band- and energy-limited signals (Q581360)

The paper is concerned with the recovery of band- and energy-limited signals from a finite number of samples taken in a given finite interval. It is proven that if m(\(\epsilon)\) is the number of samples required to obtain an \(\epsilon\)-accurate approximation of a signal from the above class, the following relation holds: \[ \lim_{\epsilon \to 0^+}\frac{m(\epsilon)\log \log (1/\epsilon)}{\log (1/\epsilon)}=1. \] It is also shown that Lagrangian interpolation with \(m(\epsilon)(1+O(1))\) arbitrary nodes yields, for sufficiently small \(\epsilon >0\), an \(\epsilon\)-approximation with almost minimal cost.