An Integer Program for Pricing Support Points of Exact Barycenters

TL;DR

This paper introduces an integer programming approach to compute exact barycenters of discrete measures, improving scalability and solution quality over traditional methods by combining linear and integer programming techniques.

Contribution

It develops a novel mixed-integer programming model that enhances the approximation of exact barycenters, addressing scalability issues in high-dimensional cases.

Findings

The method improves computational efficiency for barycenter problems.

The approach yields more accurate barycenter approximations.

Practical results demonstrate the effectiveness of the tailored branch-and-bound routine.

Abstract

The computation of exact barycenters for a set of discrete measures is of interest in applications where sparse solutions are desired, and to assess the quality of solutions returned by approximate algorithms and heuristics. The task is known to be NP-hard for growing dimension and, even in low dimensions, extremely challenging in practice due to an exponential scaling of the linear programming formulations associated with the search for sparse solutions. A common approach to facilitate practical computations is an approximation based on the choice of a small, fixed set $S_0$ of support points, or a fixed set $S^*_0$ of combinations of support points from the measures, that may be assigned mass. Through a combination of linear and integer programming techniques, we model an integer program to compute additional combinations, and in turn support points, that, when added to $S^*_0$ or $S_0$, allow for a better approximation of the underlying exact barycenter problem. The approach improves on the scalability of a classical column generation approach: instead of a pricing problem that has to evaluate exponentially many reduced cost values, we solve a mixed-integer program of quadratic size. The properties of the model, and practical computations, reveal a tailored branch-and-bound routine as a good solution strategy.