Relaxation methods for the Generalized Minimum Spanning Tree Problem

Given an undirected graph whose nodes are partitioned into a number of clusters, the Generalized Minimum Spanning Tree problem (denoted GMSTP) is then to find a minimum-cost tree which includes exactly one node from each cluster. We present several integer programming formulations of the problem and compare the polyhedra defined by the LP relaxations of these formulations. Based on a new formulation of the GMSTP we give an heuristic solution, a lower bound procedure and discuss the advantages of our approach in comparison with an erlier method.