Convex Integer Optimization by Constantly Many Linear Counterparts
Abstract
In this article we study convex integer maximization problems with composite objective functions of the form , where is a convex function on and is a matrix with small or binary entries, over finite sets of integer points presented by an oracle or by linear inequalities. Continuing the line of research advanced by Uri Rothblum and his colleagues on edge-directions, we introduce here the notion of {\em edge complexity} of , and use it to establish polynomial and constant upper bounds on the number of vertices of the projection and on the number of linear optimization counterparts needed to solve the above convex problem. Two typical consequences are the following. First, for any , there is a constant such that the maximum number of vertices of the projection of any matroid by any binary matrix is regardless of and ; and the convex matroid problem reduces to greedily solvable linear counterparts. In particular, . Second, for any , there is a constant such that the maximum number of vertices of the projection of any three-index transportation polytope for any by any binary matrix is ; and the convex three-index transportation problem reduces to linear counterparts solvable in polynomial time.
Cite
@article{arxiv.1208.5639,
title = {Convex Integer Optimization by Constantly Many Linear Counterparts},
author = {Shmuel Onn and Michal Rozenblit},
journal= {arXiv preprint arXiv:1208.5639},
year = {2014}
}