Improved Adaptive and Multi-group Parallel Genetic Algorithm Based on Good-point Set

This paper puts forward an adaptive genetic algorithm to solve the multi-group homogenization in the solution space. The use of good-point set approach improves the initial population, ensuring them a uniform distribution in the solution space. In the evolution, each population implements independent genetic operations (selection, good-point set crossover, and mutation). The introduction of adaptive operator makes crossover and mutation operator self-adaptive. As the algorithm adopts a strategy of retaining the best, a space compression strategy can be designed based on information entropy theory through the information of all sub-populations in the evolution process, which ensures the algorithmic stability and fast convergence to the global optimal solution. Furthermore, in order to explore the feasibility and effectiveness of the improved multi-group parallel algorithm, optimization tests are implemented on some of the typical multi-peak functions, and the results are compared with the analytic solution and optimal solution of basic GA. The outcome suggests that the global searching ability and convergence of the improved algorithm is far better than the basic one.