A novel harmony search algorithm with gaussian mutation for multi-objective optimization

This paper proposes a novel harmony search algorithm with gaussian mutation (GMHS) for multi-objective optimization. Harmony search (HS) algorithm has shown many advantages in solving global optimization problems; however, it also has some shortcomings, such as poor ability of escaping from local optimum and poor convergence. In view of the weaknesses of HS algorithm, several important improvements are employed in the proposed GMHS, including (a) the memory consideration rule is modified to improve convergence, (b) two bandwidths in pitch adjustment are designed to obtain better exploration and exploitation, (c) chaotic maps are utilized in the GMHS to enhance global search ability and (d) a gaussian mutation operator is employed in the GMHS to speed up convergence rate and to jump out the local optimum. To solve multi-objective optimization problems, the GMHS uses fast non-dominated sorting and crowding distance method to update harmony memory. For the purpose of preserving non-dominated solutions found during the entire search process, an external archive has been adopted. To demonstrate the effectiveness of the GMHS, it is tested with benchmark problems. The experimental results show that the GMHS is competitive in convergence and diversity performance, compared with other multi-objective evolutionary algorithms. Finally, the impact of two key parameters on the performance of GMHS is also analyzed.