Enhanced Gaussian Bare-Bone Imperialist Competition Algorithm Based on Doubling Sampling and Quasi-oppositional Learning for Global Optimization

Abstract

Gaussian bare-bone imperialist competitive algorithm (GBB-ICA) is an effective variant of imperialist competitive algorithm (ICA), which updates the position of colonies by sampling a Gaussian distribution. However, the mean and standard deviation adopted by GBB-ICA is calculated only using the positions of imperialist and the colony itself, making the searching tends to trap into local optimum. To overcome this drawback, a new double Gaussian sampling strategy is proposed in this paper. An extra Gaussian sampling point, whose mean and standard is calculated using the positions of the second best colony and the current colony itself, is introduced into GBB-ICA. To further speed up the convergence and explore informative region, the quasi-oppositional learning technique is incorporated into GBB-ICA to produce more potential candidates in the assimilation step as well as generating a higher quality initial population. The proposed algorithm is called quasi-oppositional learning-based double Gaussian sampling bare-bone imperialist competitive algorithm (QOLBDGSBB-ICA) and is tested on 20 benchmark functions and four engineering design problems. Experimental results show that the proposed algorithm outperforms over other referenced ICA variants on 19 benchmark functions, which well validates the effectiveness of the proposed algorithm.