Affiliation:
1. Amsterdam Business School, University of Amsterdam, 1001 NL Amsterdam, The Netherlands;
2. Management Science and Operations, London Business School, London NW1 4SA, United Kingdom;
3. Industrial Engineering and Operations Research, Columbia University, New York, New York 10027
Abstract
Properly defining the center of a set has been a longstanding question in applied mathematics, with implications in numerical geometry, physics, and optimization algorithms. Minkowski centers are one such definition, whose theoretical benefits are numerous and well documented. In this paper, we revisit the advantages of Minkowski centers from a computational, rather than theoretical, perspective. First, we show that Minkowski centers are solutions to a robust optimization problem. Under this lens, we then provide computationally tractable reformulations or approximations for a series of sets, including polyhedra, polyhedral projections, and intersections of ellipsoids. Computationally, we illustrate that Minkowski centers are viable alternatives to other centers, such as Chebyshev or analytic centers, and can speed up the convergence of numerical algorithms like hit-and-run and cutting-plane methods. We hope our work sheds new and practical light on Minkowski centers and exposes their potential benefits as a computational tool.
Publisher
Institute for Operations Research and the Management Sciences (INFORMS)
Subject
Management Science and Operations Research,Computer Science Applications