| 摘 要: 针对NSGA-Ⅱ算法在高维多目标优化时选择压力较小,不适用于高维空间的问题,提出一种基于简化超体积的NSGA-Ⅱ算法,利用超体积在高维空间中可以准确评价个体优劣的特点,使用简化超体积代替拥挤距离对种群中的个体进行比较,在更新种群时保留收敛性和分布性更好的个体。通过与4个先进的、具有代表性的高维多目标进化算法(NSGA-Ⅲ、MOEA/DD、KnEA、RVEA)的对比实验表明,基于简化超体积的NSGA-Ⅱ算法在求解大多数测试函数时,获得了更优的解集,证明了该算法处理高维多目标优化问题的优越性能。 |
| 关键词: 高维多目标优化;进化算法;超体积 |
|
中图分类号: TP301
文献标识码: A
|
|
| NSGA-Ⅱ Algorithm Based on Simplified Hypervolume |
|
JI Hong, ZHAO Jianyin, CHEN Jian, GE Rui
|
(Naval Aviation University, Yantai 264001, China)
ytjihong@163.com; 13791182798@163.com; 57991949@qq.com; gr33995@126.com
|
| Abstract: Aiming at the problem that the NSGA-Ⅱ algorithm has low selection pressure in many-objective optimization and is not suitable for high-dimensional space, this paper proposes a NSGA-Ⅱ algorithm based on simplified hypervolume. As hypervolume can accurately evaluate the advantages and disadvantages of individuals in high-dimensional space, individuals in the population are compared by simplified hypervolume instead of crowding distance, and individuals with better convergence and distribution are retained when updating the population. The comparative experiment with four many-objective evolutionary algorithms (NSGA-Ⅲ, MOEA/DD, KnEA, RVEA) shows that the proposed NSGA-Ⅱ algorithm based on simplified hypervolume achieves a better solution set when solving most test functions, which proves its excellent performance in handling many-objective optimization problems. |
| Keywords: many-objective optimization; evolutionary algorithm; hypervolume |
