Competitive teaching–learning-based optimization for multimodal optimization problems.

Teaching–learning-based optimization is one of the latest metaheuristic algorithms. TLBO has a simple framework and good global search ability. In addition, TLBO only needs population size and terminal condition for performing search tasks. Given these advantages, TLBO has been used widely since it was proposed. However, TLBO may fall into local optimal solutions in solving complex multimodal optimization problems. This paper reports an improved TLBO, namely competitive teaching–learning-based optimization, for solving multimodal optimization problems. In CTLBO, population is first divided into outstanding group and common group by the designed competitive mechanism. Then outstanding group is updated by the learning strategies of TLBO and common group is guided by outstanding group. In addition, a mutation operator for the optimal individual is introduced to increase the ability of CTLBO to escape from the local optima. The performance of CTLBO is investigated by 45 benchmark test functions from CEC 2014 and CEC 2015 test suites and three challenging real-world engineering problems. Experimental results show that CTLBO is more reliable and efficient on most test cases than TLBO and the other compared algorithms. This supports the effectiveness of the improved strategies and the superiority of CTLBO in solving multimodal optimization problems. [ABSTRACT FROM AUTHOR]