Parametrizing an integer linear program by an integer

TL;DR

This paper proves that the _\u03bb(t) functions, representing the _rgest values of a parameterized integer linear program, are eventually quasi-polynomial, extending previous results to broader conditions.

Contribution

It establishes that the _rgest value functions are eventually quasi-polynomial for all ll, removing previous size restrictions on the vertices of the convex hull.

Findings

_rgest value functions are eventually quasi-polynomial.

The vertices of the convex hull of lattice points have this structure without size restrictions.

Extends prior work by Calegari and Walker to more general convex hulls.

Abstract

We consider a family of integer linear programs in which the coefficients of the constraints and objective function are polynomials of an integer parameter $t.$ For $\ell$ in $\mathbb{Z}_+,$ we define $f_\ell(t)$ to be the $\ell^{\text{th}}$ largest value of the objective function with multiplicity for the integer linear program at $t.$ We prove that for all $\ell,$ $f_\ell$ is eventually quasi-polynomial; that is, there exists $d$ and polynomials $P_0, \ldots, P_{d-1}$ such that for sufficiently large $t,$ $f_\ell(t)=P_{d \pmod{t}}(t).$ Closely related to finding the $\ell^{\text{th}}$ largest value is describing the vertices of the convex hull of the feasible set. Calegari and Walker showed that if $R(t)$ is the convex hull of $\mathbf{v_1}(t), \ldots, \mathbf{v_k}(t)$ where $\mathbf{v_i}$ is a vector whose coordinates are in $\mathbb{Q}(u)$ and of size $O(u),$ then the vertices of the convex hull of the set of lattice points in $R(t)$ has eventually quasi-polynomial structure. We prove this without the $O(u)$ assumption.