Incorporating Great Deluge with Harmony Search for Global Optimization Problems

Abstract

Harmony search (HS) algorithm is relatively a recent metaheuristic optimization method inspired by natural phenomenon of musical improvisation process. Despite its success, the main drawback of harmony search are contained in its tendency to converge prematurely due to its greedy selection method. This probably leads the harmony search algorithm to get stuck in local optima and unsought solutions owing to the limited exploration of the search space. The great deluge algorithm is a local search-based approach that has an efficient capability of increasing diversity and avoiding the local optima. This capability comes from its flexible method of accepting the new constructed solution. The aim of this research is to propose and evaluate a new variant of HS. To do so, the acceptance method of the great deluge algorithm is incorporated in the harmony search to enhance its convergence properties by maintaining a higher rate of diversification at the initial stage of the search process. The proposed method is called Harmony Search Great Deluge (HS-GD) algorithm. The performance of HS-GD and the classical harmony search algorithm was evaluated using a set of ten benchmark global optimization functions. In addition, five benchmark functions of the former set were employed to compare the results of the proposed method with three previous harmony search variations including the classical harmony search. The results show that HS-GD often outperforms the other comparative approaches.