Teaching–learning-based genetic algorithm (TLBGA): an improved solution method for continuous optimization problems.

The mutation is one of the most important stages in the genetic algorithm (GA) because of its influence on exploring solution space and overcoming premature convergence. Since there are many types of mutation operators, the problem lies in selecting the appropriate type, and so, researchers usually need more trial and error. This paper investigates a new mutation operator based on teaching–learning-based optimization (TLBO) to enhance the performance of genetic algorithms. This new mutation operator treats intelligently instead of the random type, enhances the quality of solution, and speeds up the convergence of GA simultaneously. Several experiments are conducted on six standard test functions to evaluate the effect of the proposed mutation operator. First, proper comparisons are made between the performance of the proposed mutation to the classic mutation of GA and their combinatorial format. The result indicates the effect of the proposed mutation operator on the significant enhancement of the genetic algorithms' performance particularly. Due to computational analysis with Intel(R) Core(TM) i5-2430 M CPU @ 2.40 GHz processor, this method causes 32–53.3% reduction in essential iteration to present zero amount as the final value for four test functions (i.e., Beale, Himmelblau, Booth, and Rastrigin). For the two other functions that provides a non-zero value (i.e., Ackley and Sphere), the proposed method improves nearly 100% in average of objective. According to the result, the final solutions of the proposed method are equal or better than the classic GA in all six problems. Then, the performance of the proposed algorithm in comparison to five well-known algorithms ensures its superiority. In all comparisons, the proposed method performs equal or better than five other algorithms in CPU time and quality solutions. [ABSTRACT FROM AUTHOR]