The vector partition problem for convex objective functions

The vector partition problem for convex objective functions

0.00 Avg rating0 Votes
Article ID: iaor20041222
Country: United States
Volume: 26
Issue: 3
Start Page Number: 583
End Page Number: 590
Publication Date: Aug 2001
Journal: Mathematics of Operations Research
Authors: ,
Abstract:

The partition problem concerns the partitioning of a given set of n vectors in d-space into p parts to maximize an objective function that is convex on the sum of vectors in each part. The problem has broad expressive power and captures NP-hard problems even if either p or d is fixed. In this article we show that when both p, d are fixed, the problem is solvable in strongly polynomial time using O(nd(p−1)−1) arithmetic operations. This improves upon the previously known bound of O(ndp2). Our method is based on the introduction of the signing zonotope of a set of points in space. We study this object, which is of interest in its own right, and show that it is a refinement of the so-called partition polytope of the same set of points.

Reviews

Required fields are marked *. Your email address will not be published.