A Comparative Assessment of Memetic, Evolutionary, and Constructive Algorithms for the Multiobjective d-MST Problem

Finding a minimum-weight spanning tree (MST) in a graph is a classic problem in operational research with important applications in network design. In this paper, we consider the degree-constrained multi-objective MST problem, which is NP-hard. On fteen benchmark instances, we compare the performance of three diierent algorithms: the Pareto archived evolution strategy (PAES); a new multiobjective evolutionary algorithm, AESSEA; and the memetic PAES algorithm, M-PAES, all employing the same initialization procedure, encoding and operators. We nd M-PAES performs well on the whole range of problem types, generally outperforming the pure evolutionary and local search algorithms.