Nonuniform sampling and recovery of multidimensional bandlimited functions by Gaussian radial-basis functions (Q719690)

Let \(d\geq 1\) be a natural number and consider a convex and symmetric (about the origin) set \(Z\subset {\mathbb R}^d\); denote by \(L_2(Z)\) the space of all complex-valued, square integrable functions (with respect to the Lebesgue measure) on \(Z\). The sequence \((x_j:j\in{\mathbb N})\) is a Riesz-basis sequence for \(L_2(Z)\) (i.e. the sequence \((e^{-i\langle x_j, \cdot\rangle}:j\in {\mathbb N})\) of exponential functions is a Riesz basis for \(L_2(z)\)). Finally, let for an arbitrary bounded subset \(S\in {\mathbb R}^d\), \(PW_S\) (the \textit{Paley-Wiener space} associated to \(S\)) be the space of all square integrable functions on \({\mathbb R}^d\) whose Fourier tranforms vanish outside \(S\). The authors then first introduce for \(Z\) as above, with \(m(Z)>0\), and a given Riesz-basis sequence, the \textit{Gaussian interpolation operator} as follows. Let be given a fixed positive number \(\lambda\). For any \(f\in PW_Z\) there exists a unique square-summable sequence \((a_j^{(\lambda)}\,:\,j\in{\mathbb N})\) such that \[ \sum_{j\in {\mathbb N}}\,a_j^{(\lambda)}g_{\lambda}(x_k-x_j)=f(x_k),\;k\in{\mathbb N}. \] The Gaussian interpolation operator \(I_{\lambda}\,:\,PW_Z\rightarrow L_2( {\mathbb R}^d)\) is then defined by \[ I_{\lambda}(f)(\cdot)=\sum_{j\in{\mathbb N}}\,a_j^{(\lambda)}g_{\lambda}((\cdot)-x_j). \] This is a well-defined, bounded linear operator, moreover \(I_{\lambda}(f)\in C_0( {\mathbb R}^d)\) (the continuous functions with \(\lim_{\|x\|_2\rightarrow\infty}\,f(x)=0\)). The main result of the paper is then {Theorem 3.6}: Let \(\delta\in (\sqrt{2/3},1]\) and \(0<\beta <\sqrt{3\delta^2-2}\). Assume that \(Z\subset {\mathbb R}^d\) with Riesz-basis sequence is given as above and satisfies \(\delta B_2\subset Z\subset B_2\). Then, for every \(f\in PW_{\beta B_2}\) we have \[ f=\lim_{\lambda\downarrow 0}\,I_{\lambda}(f) \] in \(L_2({\mathbb R}^d)\) and uniformly on \( {\mathbb R}^d\). The authors forget, however, to introduce the set \(B_2\) explicitly; they probably use it to indicate the unit-ball (with respect to Cartesian coordinates) in \({\mathbb R}^d\).