The Kramer sampling theorem revisited (Q2255601)

Kramer's sampling theorem states that, if there exists a function \(K(x,t)\) continuous in \(t\) such that \(K(x,t) \in L^2(I)\) as a function of \(x\) for every real number \(t\) and there exists a sequence of real numbers \(\{t_n\}_{n\in \mathbb{Z}}\) such that \(\{K(x,t_n)\}_{n \in \mathbb{Z}}\) is a complete orthogonal family in \(L^2(I)\) for some finite interval \(I=[a,b]\), then for any function of the form \[ f(t)=\int_a^b \overline{F}(x) {K}(x,t)\, dx=\langle F, K\rangle , \] with \(F \in L^2(I)\), we have \[ f(t)= \sum_{n=-\infty}^{\infty} f(t_n) S_n^{\ast}(t) , \] where \[ S_n^{\ast}(t)=\frac{\int_a^b {K}(x,t)\overline{K}(x,t_n)dx}{\int_a^b|K(x,t_n)|^2 dx} . \] The theorem is called a sampling theorem because it admits the reconstruction of the function \(f\) from its samples at the points \(\{t_n\}_{n\in \mathbb{Z}}\). In addition to being a generalization of the Whittaker-Shannon-Kotel'nikov sampling theorem, Kramer's theorem establishes a link between sampling theorems and boundary value problems and offers ground for abstract generalizations. The novelty of Kramer's work is not so much in the statment of his theorem as in his observation of how to generate the kernel function \(K(x,t)\) and the sampling points \(\{t_n\}_{n\in \mathbb Z}\). Kramer noticed that the kernel function and the sampling points could be found from certain boundary value problems. The article under review is a survey article on Kramer's sampling theorem and some of its generalizations. The authors survey their own results and contribution to the subject. The different generalizations of Kramer's theorem discussed are variations of the main theorem that was derived by the authors in an earlier work and which gives a Kramer-type sampling expansion in the setting of reproducing kernel Hilbert spaces as developed by \textit{S. Saitoh} [Integral transforms, reproducing kernels and their applications. Harlow: Longman (1997; Zbl 0891.44001)]. Examples of some of the generalizations discussed are sampling in abstract reproducing kernel Hilbert spaces, Kramer's sampling theorem and indeterminate moment problems, sampling in shift-invariant spaces, and sampling in reproducing kernel Hilbert spaces of entire functions. The authors also discuss conditions under which a Kramer sampling series is a Lagrange interpolation series. Although reproducing kernel spaces were originally introduced in the Hilbert space setting, currently there are some extensions to Banach spaces. The authors discuss one of these extensions and derive a Kramer-type sampling theorem in a reproducing kernel Banach space. They, then proceed to derive a Kramer-type sampling theorem in a reproducin kernel space of distributions.