Strengthened semidefinite relaxations via a second lifting for the Max-Cut problem
Strengthened semidefinite relaxations via a second lifting for the Max-Cut problem
In this paper we study two strengthened semidefinite programming relaxations for the Max-Cut problem. Our results hold for every instance of Max-Cut; in particular, we make no assumptions about the edge weights. We prove that the first relaxation provides a strengthening of the Goemans-Williamson relaxation. The second relaxation is a further tightening of the first one and we prove that its feasible set corresponds to a convex set that is larger than the cut polytope but nonetheless is strictly contained in the intersection of the elliptope and the metric polytope. Both relaxations are obtained using Lagrangian relaxation. Hence our results also exemplify the strength and flexibility of Lagrangian relaxation for obtaining a variety of SDP relaxations with different properties. We also address some practical issues in the solution of these SDP relaxations. Because Slater's constraint qualification fails for both of them, we project their feasible sets onto a lower dimensional space in a way that does not affect the sparsity of these relaxations but guarantees Slater's condition. Some preliminary numerical results are included.
Max-Cut problem, semidefinite programming relaxations, Lagrangian relaxation, Cut polytope metric polytope
79-106
Anjos, Miguel F.
0103b24d-cab5-4109-9924-dc9eebd04257
Wolkowicz, Henry
3ef6bc24-0956-4eb2-a542-802d82454cf7
2002
Anjos, Miguel F.
0103b24d-cab5-4109-9924-dc9eebd04257
Wolkowicz, Henry
3ef6bc24-0956-4eb2-a542-802d82454cf7
Anjos, Miguel F. and Wolkowicz, Henry
(2002)
Strengthened semidefinite relaxations via a second lifting for the Max-Cut problem.
Discrete Applied Mathematics, 119 (1-2), .
(doi:10.1016/S0166-218X(01)00266-9).
Abstract
In this paper we study two strengthened semidefinite programming relaxations for the Max-Cut problem. Our results hold for every instance of Max-Cut; in particular, we make no assumptions about the edge weights. We prove that the first relaxation provides a strengthening of the Goemans-Williamson relaxation. The second relaxation is a further tightening of the first one and we prove that its feasible set corresponds to a convex set that is larger than the cut polytope but nonetheless is strictly contained in the intersection of the elliptope and the metric polytope. Both relaxations are obtained using Lagrangian relaxation. Hence our results also exemplify the strength and flexibility of Lagrangian relaxation for obtaining a variety of SDP relaxations with different properties. We also address some practical issues in the solution of these SDP relaxations. Because Slater's constraint qualification fails for both of them, we project their feasible sets onto a lower dimensional space in a way that does not affect the sparsity of these relaxations but guarantees Slater's condition. Some preliminary numerical results are included.
This record has no associated files available for download.
More information
Published date: 2002
Keywords:
Max-Cut problem, semidefinite programming relaxations, Lagrangian relaxation, Cut polytope metric polytope
Organisations:
Operational Research
Identifiers
Local EPrints ID: 29672
URI: http://eprints.soton.ac.uk/id/eprint/29672
ISSN: 0166-218X
PURE UUID: 917ec2d4-20c7-4616-ada6-a8b4ff3944cf
Catalogue record
Date deposited: 12 May 2006
Last modified: 15 Mar 2024 07:33
Export record
Altmetrics
Contributors
Author:
Miguel F. Anjos
Author:
Henry Wolkowicz
Download statistics
Downloads from ePrints over the past year. Other digital versions may also be available to download e.g. from the publisher's website.
View more statistics