Team, Visitors, External Collaborators
Overall Objectives
Research Program
Application Domains
Highlights of the Year
New Software and Platforms
New Results
Bilateral Contracts and Grants with Industry
Partnerships and Cooperations
XML PDF e-pub
PDF e-Pub

Section: Research Program

Polyhedral approaches for MIP

Adding valid inequalities to the polyhedral description of an MIP allows one to improve the resulting LP bound and hence to better prune the enumeration tree. In a cutting plane procedure, one attempt to identify valid inequalities that are violated by the LP solution of the current formulation and adds them to the formulation. This can be done at each node of the branch-and-bound tree giving rise to a so-called branch-and-cut algorithm [65]. The goal is to reduce the resolution of an integer program to that of a linear program by deriving a linear description of the convex hull of the feasible solutions. Polyhedral theory tells us that if X is a mixed integer program: X=Pn×p where P={xn+p:Axb} with matrix (A,b)m×(n+p+1), then conv(X) is a polyhedron that can be described in terms of linear constraints, i.e. it writes as conv(X)={xn+p:Cxd} for some matrix (C,d)m'×(n+p+1) although the dimension m' is typically quite large. A fundamental result in this field is the equivalence of complexity between solving the combinatorial optimization problem min{cx:xX} and solving the separation problem over the associated polyhedron conv(X): if x˜conv(X), find a linear inequality πxπ0 satisfied by all points in conv(X) but violated by x˜. Hence, for NP-hard problems, one can not hope to get a compact description of conv(X) nor a polynomial time exact separation routine. Polyhedral studies focus on identifying some of the inequalities that are involved in the polyhedral description of conv(X) and derive efficient separation procedures (cutting plane generation). Only a subset of the inequalities Cxd can offer a good approximation, that combined with a branch-and-bound enumeration techniques permits to solve the problem. Using cutting plane algorithm at each node of the branch-and-bound tree, gives rise to the algorithm called branch-and-cut.