Tagging Metaheuristics with Problem-Oriented Labels for Non-Expert Users

Ye Tian, Maihai Sun, Shangshang Yang, Xingyi Zhang

Abstract: Represented by evolutionary computation and swarm intelligence, nature-inspired metaheuristics have shown superiority in solving complex optimization problems for decades. In spite of the source codes of state-of-the-art metaheuristics provided in existing libraries such as PlatEMO, it is still difficult for researchers unfamiliar with metaheuristics to select a suitable one for solving their problems. Since no metaheuristic can per-form the best on all problems, selecting a suitable metaheuristic is very critical for handling real-world applications. Therefore, this work suggests a set of problem-oriented labels for tagging metaheuristics, where metaheuristics are not simply classified into disjoint categories but tagged with multiple labels. By referring to the labels of a metaheuristic, researchers can easily know whether it is capable of solving a given type of problems. The proposed labels can highly facilitate the selection of metaheuristics.