Enhanced self-adaptive global-best harmony search without any extra statistic and external archive

Harmony search is a music-inspired optimization algorithm. Our findings reveal that the common uniform randomization selection in the harmony search algorithm is less efficient than the Gaussian mutation to explore the global optimal solution, regardless of solving unimodal or multimodal problems under the same parameters settings. To evaluate the effectiveness of a given search strategy, a general measurement that can apply to other algorithms is also proposed. To enhance the search efficiency and effectiveness, a self-adaptive global-best harmony search algorithm is developed. The proposed algorithm takes full advantage of the valuable information hidden in harmony memory to devise a high-performance search strategy and integrates a self-adaptive mechanism to develop a parameter-setting-free technique. Moreover, it is as simple and straightforward to implement as the canonical harmony search algorithm. It does not require any extra statistic and external archive. It well maintains the interesting and distinctive framework of the original version in the evolutionary computation domain. The experimental results show that the proposed algorithm significantly outperforms the recent adaptive variant of the harmony search algorithm and achieves the strongly competitive performances compared with other state-of-the-art adaptive evolutionary algorithms. The proposed algorithm is also successfully applied into the real-world space trajectory optimization problem.