A Simple Brain Storm Optimization Algorithm With a Periodic Quantum Learning Strategy

Brain storm optimization (BSO) is a young and promising population-based swarm intelligence algorithm inspired by the human process of brainstorming. The BSO algorithm has been successfully applied to both science and engineering issues. However, thus far, most BSO algorithms are prone to fall into local optima when solving complicated optimization problems. In addition, these algorithms adopt complicated clustering strategies, such as K-means clustering, resulting in large computational burdens. This paper proposes a simple BSO algorithm with a periodic quantum learning strategy (SBSO-PQLS), which includes three new strategies developed to improve the defects described above. First, we develop a simple individual clustering strategy that sorts individuals according to their fitness values and then allocates all individuals into different clusters. This reduces computational burdens and resists premature convergence. Second, we present a simple individual updating strategy by simplifying the individual combinations and improving the step size function to enrich the diversity of newly generated individuals and reduce redundancy in the pattern for generating individuals. Third, a quantum-behaved individual updating with periodic learning (QBIU-PL) strategy is developed by introducing a quantum-behaved mechanism into SBSO-PQLS. QBIU-PL provides new momentum, enabling individuals to escape local optima. With the support of these three strategies, SBSO-PQLS effectively improves its global search capability and computational burdens. SBSO-PQLS is compared with seven other BSO variants, particle swarm optimization, and differential evolution on CEC2013 benchmark functions. The results show that SBSO-PQLS achieves a better global search performance than do the other nine algorithms.