Closures of disjunctive relaxations and valid inequalities for mixed integer problems and applications
Résumé
Valid inequalities deduced from the elementary closure of Lift-and-Project cuts have provided an efficient way of computing bounds of good quality for some classes of mixed integer programming problems such as the maximum 2-satisfiability problem (MAX-2SAT), as shown by Bonami and Minoux (2006). In the present talk, we will first recall how the structure of the optimization problem over this elementary closure can be exploited to generate, in an efficient way, strong valid inequalities. We will then discuss connections with other types of relaxations (such as Sherali-Adams), together with possible extensions of the approach using higher rank disjunctive relaxations. Some preliminary computational results will be shown to illustrate the potential interest and the difficulties related to the practical use of such relaxations.