Steady-State Brain Storm Optimization Using Double Roulette Selection

Zhenshou Song, Handing Wang

Abstract: Brain storm optimization (BSO) is a newly proposed bionic population-based optimization algorithm. It has successfully handled a variety of optimization problems in many fields. However, most existing BSO variants rarely focus on the selection method of BSO. The original BSO selection method is oversimplified, which requires a large number of evaluations. To address this issue, we develop a steady-state BSO with a double roulette selection mechanism (SSBSO-DRS). First, a double roulette selection mechanism is proposed, thoroughly considering ranking index information among solutions. Second, we design an improved solution clustering strategy that enables each cluster to have the same quality of solutions, which resists premature convergence and reduces computational burdens. Third, considering the interaction between the same round of solutions in a real brainstorming process, a steady-state selection model is introduced to replace a generational selection model. SSBSO-DRS is compared with six popular BSO algorithms on the CEC2013 test suite. Experimental results show that SSBSODRS obtains the competitive performance on the global search capability.