Bernard Gendron (University of Montréal)
Decomposition Methods for Network Design
Network design applications are prevalent in telecommunications, transportation and logistics. We consider multicommodity network design problems that capture three important features of network design applications: the interplay between investment and operational costs, the multicommodity aspect, and the presence of capacity constraints. We focus on mathematical programming approaches and present methods designed to solve large-scale network design instances: cutting-plane and column generation methods, as well as Benders decomposition and Lagrangian relaxation approaches.