Population Diversity Guided Dimension Perturbation for Artificial Bee Colony Algorithm

Abstract

In the original artificial bee colony (ABC), only one dimension of the solution is updated each time and this leads to little differences between the offspring and the parent solution. Then, it affects the convergence speed. In order to accelerate the convergence speed, we can update multiple dimensions of the solution at the same time, and the information of the global optimal solution can be used for guidance. However, using these two methods will reduce the population diversity at the initial stage. This is not conducive to search of multimodal functions. In this paper, a population diversity guided dimension perturbation for artificial bee colony algorithm (called PDDPABC) is proposed, in which population diversity is used to control the number of dimension perturbations. Then, it can maintain a certain population diversity, and does not affect the convergence speed. In order to verify the performance of PDDPABC, we tested its performance on 22 classic problems and CEC 2013 benchmark set. Compared with several other ABC variants, our approach can achieve better results.