摘要
针对传统的非支配排序在处理高维多目标优化问题过程中,因非支配解数量的指数增长而导致算法没有足够的选择压力的情况,提出了一种知识驱动的高维多目标算法(KD-NSGA-Ⅲ),以提升算法的收敛速度.该算法采用Pareto支配关系来衡量种群的优劣,通过知识引导筛选优势种群,结合参考点选择机制增强种群的多样性和广泛性,同时采用模拟二进制交叉策略和改进的自适应变异策略增强算法的搜索能力.通过对标准测试函数的实验,结果表明:相较于一些经典的多目标优化算法,KD-NSGA-Ⅲ在高维多目标优化问题方面性能改善效果优异,尤其在收敛速度上有显著的提升.
In order to solve the problem of the selection pressure loss in traditional non-dominated method due to the exponential growth of the amount of non-dominated solutions in many-objective optimization problems,a knowledge-driven high-dimensional many-objective algorithm(KD-NSGA-Ⅲ) which improves the convergence speed of the algorithm was proposed.In the proposed KD-NSGA-Ⅲ,Pareto dominance was used to measure the quality of the population,dominant populations were filtered through guidance of knowledge and the reference point selection mechanism was combined to enhance the diversity and extensiveness of the population.In the meantime,simulated binary crossover strategy and improved adaptive mutation strategy were used to enhance the ability to search for the algorithm.The experiments of standard test functions show that KD-NSGA-Ⅲ has excellent performance in many-objective optimization problems,especially in convergence speed compared with some classical many-objective optimization algorithms.
作者
胡成玉
余果
代立国
颜雪松
HU Chengyu;YU Guo;DAI Liguo;YAN Xuesong(School of Computer Science,China University of Geosciences(Wuhan),Wuhan 430074,China)
出处
《华中科技大学学报(自然科学版)》
EI
CAS
CSCD
北大核心
2020年第6期19-25,共7页
Journal of Huazhong University of Science and Technology(Natural Science Edition)
基金
国家自然科学基金资助项目(U1911205)
华中科技大学数字制造装备与技术国家重点实验室开放基金资助项目(DMETKF2019018)。
关键词
知识驱动
高维多目标优化
自适应变异
非支配排序
收敛速度
knowledge-driven
many-objective optimization
adaptive mutation
non-dominated solutions
convergence speed