Bounds on the number of integer points in a polytope via concentration estimates

Abstract

It is generally hard to count, or even estimate, how many integer points lie in a polytope P. Barvinok and Hartigan have approached the problem by way of information theory, showing how to efficiently compute a random vector which samples the integer points of P with (computable) constant mass, but which may also land outside P. Thus, to count the integer points of P, it suffices to determine the frequency with which the random vector falls in P. We prove a collection of efficiently computable upper bounds on this frequency. We also show that if P is suitably presented by n linear inequalities and m linear equations (m fixed), then under mild conditions separating the expected value of the above random vector from the origin, the frequency with which it falls in P is O(n^{-m/2}) as n -> infinity. As in the classical Littlewood-Offord problem, all results in the paper are obtained by bounding the point concentration of a sum of independent random variables; we sketch connections to previous work on the subject.