Random sampling of bandlimited functions (Q611016)

This paper considers the problem of random sampling for bandlimited functions, that is, randomly chosen samples \(f(x_j), j\in J\subset \mathbb{N}\) so that a sampling inequality \(A\|f\|_2^2 \leq \sum_{j\in J} |f(x_j)|^2\leq B\|f\|_2^2\) holds uniformly for all bandlimited functions \(f\) with \(\text{supp} \hat f\subseteq [-1/2,1/2]^d\). The main results of the paper reveal some interesting phenomena on random sampling of bandlimited functions. On the one hand, the paper shows that with probability one, the sampling inequality fails for any reasonable definition of a random set on \(\mathbb{R}^d\), e.g., for uniform distribution over disjoint cubes (see Proposition~2.2) or for spatial Poisson processes (see Proposition~2.3). On the other hand, the paper shows in Theorem~3.1 that with overwhelming probability, for uniform distribution over the cube \([-R/2,R/2]^d\), the sampling inequality holds for certain compact subsets \(\mathcal{B}(R, \delta)\) (which is defined in Definition~1) of the space of bandlimited functions and for sufficiently large sampling size.