par Hansen, Pierre 
;Schindl, David;Labbé, Martine
Référence Discrete Optimization, 6, 2, page (135-147)
Publication Publié, 2009-05
;Schindl, David;Labbé, Martine Référence Discrete Optimization, 6, 2, page (135-147)
Publication Publié, 2009-05
Article révisé par les pairs
| Résumé : | We consider two (0, 1)-linear programming formulations of the graph (vertex-) coloring problem, in which variables are associated with stable sets of the input graph. The first one is a set covering formulation, where the set of vertices has to be covered by a minimum number of stable sets. The second is a set packing formulation, in which constraints express that two stable sets cannot have a common vertex, and large stable sets are preferred in the objective function. We identify facets with small coefficients for the polytopes associated with both formulations. We show by computational experiments that both formulations are about equally efficient when used in a branch-and-price algorithm. Next we propose some preprocessing, and show that it can substantially speed up the algorithm, if it is applied at each node of the enumeration tree. Finally we describe a cutting plane procedure for the set covering formulation, which often reduces the size of the enumeration tree. © 2008 Elsevier B.V. All rights reserved. |
