元胞文化算法优化仿真研究

Optimization Design and Simulation Research on Cellular Cultural Algorithm

文化算法具有双层进化空间,通过信仰空间的知识元指引种群空间的个体进化,个体在知识指引下具有进化速度快的优点,种群有收敛速度快的优点。种群在知识指引下会加速收敛,但如果没有收敛到全局最优解而是收敛到局部最优解时,将陷入局部最优解而无法跳出。针对这个问题,有两种解决思路:一是加入某种策略防止种群陷入局部最优解;二是加入跳出机制使种群有跳出局部最优解的能力。在种群空间中引入元胞遗传算法控制知识元的影响范围,减缓文化传播速度,避免种群因收敛速度过快而陷入局部最优解。并对四种元胞结构进行比较,选取摩尔邻居结构作为算法中的元胞结构。通过标准测试函数仿真研究表明,加入了元胞结构的文化算法能减缓文化传播速度,避免种群早熟收敛,显著提高算法的收敛成功率、求解精度和稳定性。

Cultural algorithm has double evolutionary spaces including belief space and population space. The knowledge of belief space can guide individual evolution of population space and accelerate individual evolution. Cultural algorithm has the advantage that the speed of convergence is fast. The population is easy to reach the premature convergence. If the population don＇t converge in global optimal solution in control of knowledge, it is difficult for cultural algorithm to escape from the local optimum. Aiming at this problem, there are two kinds of solutions ： one is to join a strategy to prevent the population converge to a local optimal solution; the other is to design a mechanism that makes the population has the ability to jump out of local optimal solution. The population space utilized cellular genet- ic algorithm to control the range of influence of knowledge, reduce the speed of culture transmission, and avoid popu- lation into local optimal solution. According to the comparison of four kinds of cellular structures, Moore neighbor- hood structure was selected as the cellular structure of algorithm. Through standard test functions, the simulation experiment research shows that the cellular cultural algorithm can slow down cultural transmission speed and avoid premature convergence, meanwhile significantly improve the success rate of convergence of the algorithm, the precision and stability.