混合精英学习的分组APO算法

A grouping Artificial Physics Optimization Algorithm Based on Elite Learning

为解决APO算法只遵循一种运动规则,过程单一,多样性较差,易使算法陷入局部最优的不足,借鉴精英学习策略,提出了分组精英学习策略对APO算法改进。该算法对种群个体进行分组,组内个体单独进化若干代,按适应值排序后选择最好的若干个体作为精英个体,精英个体即为组间个体,进行组间搜索,同时组内个体围绕各自精英个体局部精细搜索寻优,并引入反向学习和种群多样性指标动态调整各组个体的运动趋势,使个体间相似程度增大,寻找潜在的较好解,同时对组内组间不同个体遵循不同的作用力规则,有效地保持种群多样性,通过14个测试函数与APO算法比较,实验结果表明,该算法是有效的,在种群多样性与解的精度上较优。

In order to solve the problems that the APO algorithm follows only single motion rule,the process is single and the diversity is bad,and it is also easy to make the algorithm fall into the local optimum solution,a grouping elitist learning strategy is proposed to improve the APO algorithm based on elite learning strategies. The algorithm divides the population particles into groups,and particles in the group evolve several generations independently. After sorting according to the fitness value,the best particles are selected as the elite particles,and the elite particles form a group to search,and particles within the group concentrate on elite particles to search locally. The opposition-based learning and population diversity are introduced to dynamically adjust the movement trend of each group,and the degree of similarity between particles increases to find potential better solutions. At the same time,different particles in the group follow the different rules of force to maintain the diversity of the population. By comparing the 14 test functions with the APO algorithm,the experimental results show that the algorithm is effective and has better precision of the solution and better population diversity.