A Parallel Lagrangian Relaxation Algorithm for the Min-Degree Constrained Minimum Spanning Tree Problem

Abstract

AbstractGiven an edge weighted undirected graph G and a positive integer d, the Min-Degree Constrained Minimum Spanning Tree Problem (MDMST) asks for a minimum cost spanning tree of G, such that each vertex is either a leaf or has degree at least d in the tree. The strongest known MDMST lower bounds, provided by a reformulation by intersection, are very expensive to be evaluated directly, by Linear Programming solvers. Therefore, we propose a Lagrangian Relaxation algorithm for approximating them. The reformulation makes use of a large number of variables and the relaxation involves the dualization of a large number of constraints. Attempting to speed up the computation of the Lagrangian Dual bounds, we implemented a parallel Subgradient Method. We also introduced a Lagrangian heuristic based on a Local Branching algorithm. With the proposed methods, respectively 26 and 14 new best upper and lower bounds are presented.KeywordsMin-Degree Constrained Minimum Spanning Tree ProblemLagrangian RelaxationLocal BranchingParallel Programming