Brain storm optimization using a slight relaxation selection and multi-population based creating ideas ensemble

摘要

Brain storm optimization is a swarm intelligence algorithm inspired by the brainstorming process in human beings. Many researchers have paid much more attention to it, and many attempts have been made to improve it's performance. The search ability of brain storm optimization is maintained by the creating process of ideas, but it still suffers from sticking into stagnation during exploitation phase. This paper proposes a novel brain storm optimization variant, named RMBSO, in which a slight relaxation selection and multi-population based creating ideas ensemble are employed to improve the performance of brain storm optimization on global optimization problem with diverse landscapes. Firstly, the basic framework of original brain storm optimization is imbedded into multi-population based ensemble of heterogeneous but complementary creating ideas to make the algorithm jump out of stagnation with strong searching ability. Secondly, a new triangular mutation ruler and a simple partition of subpopulations are designed to better balance exploration and exploitation. Thirdly, a slight relaxation selection mechanism instead of greedy choice is first developed to keep the population's diversity. Finally, extensive experiments on the suit of CEC 2015 benchmark functions and statistical comparisons are executed. Experimental results indicate that the proposed algorithm is significantly better than, or at least comparable to the state-of-the-art brain storm optimization variants and several improved differential evolution algorithms.