A Powerful Local Search Method for Minimum Steiner Tree Problem

Boyu Yang,Weiguo Zheng
DOI: https://doi.org/10.1007/978-981-97-7241-4_4
2024-01-01
Abstract:The minimum Steiner tree problem (STP) is a well-known combinatorial optimization problem in graph theory. However, since finding an exact solution for STP is an NP-hard problem, numerous efforts have been made to identify high-quality approximate solutions. Local search algorithms are widely used for improving the quality of solutions for Steiner trees. Essentially, these algorithms search for a new substructure to replace a subtree of a given Steiner tree. Nonetheless, none of the existing local search algorithms are adept at handling difficult STP instances because they rely too heavily on limited local substructures (namely key paths) and replace paths independently of each other. In this paper, we propose two new local search strategies based on a newly defined structure termed key star. The key star structure allows us to explore a broader search space by utilizing the relation among the replacing vertices and paths. Extensive experiments over the widely used benchmarks, including both real and synthetic graphs, demonstrate that our methods outperform the state-of-the-art local search algorithms significantly.
What problem does this paper attempt to address?