求解高维昂贵多目标问题的约束型Dropout代理辅助进化算法
Constrained Dropout Surrogate-Assisted Evolutionary Algorithm for Expensive Many-Objective Problems
- 电子学报 2023年51卷第7期 页码：1859-1867
- 作者机构：
太原科技大学计算机科学与技术学院，山西太原 030024
- 作者简介：
[ "张 睿 男，1987年2月出生于山西省太原市.现为太原科技大学计算机科学与技术学院副教授、硕士生导师.主要从事智能信息处理、自动机器学习与群体智能优化. E-mail: zhangrui@tyust.edu.cn" ]
[ "白晓露（通讯作者） 男，1999年2月出生于山西省大同市. 现为太原科技大学计算机科学与技术学院硕士研究生.主要研究方向为群体智能优化算法. E-mail: bxl17335866519@163.com" ]
- 基金信息：山西省基础研究计划（自由探索类）项目(20210302123216);山西省装备数字化与故障预测工程研究中心开放课题(ZBPHM20201104);山西省研究生教育改革研究课题项目(2021YJJG244);山西省研究生教育创新项目(2021Y699)
DOI：10.12263/DZXB.20220694
中图分类号： TP273收稿：2022-06-16，
修回：2023-03-01，
纸质出版：2023-07-25
- 稿件说明：
移动端阅览
张睿,白晓露,潘理虎.求解高维昂贵多目标问题的约束型Dropout代理辅助进化算法[J].电子学报,2023,51(07):1859-1867.
ZHANG Rui,BAI Xiao-lu,PAN Li-hu.Constrained Dropout Surrogate-Assisted Evolutionary Algorithm for Expensive Many-Objective Problems[J].ACTA ELECTRONICA SINICA,2023,51(07):1859-1867.
张睿,白晓露,潘理虎.求解高维昂贵多目标问题的约束型Dropout代理辅助进化算法[J].电子学报,2023,51(07):1859-1867. DOI： 10.12263/DZXB.20220694.
ZHANG Rui,BAI Xiao-lu,PAN Li-hu.Constrained Dropout Surrogate-Assisted Evolutionary Algorithm for Expensive Many-Objective Problems[J].ACTA ELECTRONICA SINICA,2023,51(07):1859-1867. DOI： 10.12263/DZXB.20220694.
摘要
现实优化问题常需同时对多个冲突目标进行评估优化，由于评估过程多依赖大量复杂的仿真实验，从而产生计算代价昂贵的问题.代理模型辅助下的进化算法可在计算资源有限的情况下为此类问题提供有效的最优解集.但随着问题决策以及目标空间维度变高，则会衍生出诸多负面因素限制代理模型的预测精度. 对此， 提出一种基于约束型Dropout神经网络的代理辅助进化算法（Constrained Dropout Neural Network based surrogate-assisted Evolution Algorithm，CDNNEA），以约束型Dropout神经网络作为一种可扩展方案来增强代理模型在高维空间中的适用性，在模型管理部分中，构建种群个体收敛性以及多样性评判指标，自适应地选取引导代理模型更新的代表性个体.通过在DTLZ基准测试问题上进行实验，CDNNEA显示出相较其它先进算法性能表现最优或近似最优，同时将其应用于高维参数自适应优化的现实问题中.实验表明：提供的代表性解决方案投入实际应用时识别准确率均较优且模型计算量可控，验证出CDNNEA在高维昂贵类优化问题上的有效性.
Abstract
Realistic optimization problems often need to evaluate and optimize multiple conflicting targets at the same time. Because the evaluation process depends on a large number of complex simulation experiments
the calculation cost is expensive. The evolutionary algorithms assisted by the surrogate models can provide effective optimal solution sets for such problems with limited computational resources. However
as the dimensions of decision space and target space become higher
many negative factors will be derived to limit the prediction accuracy of the surrogate model. Therefore
a surrogate-assisted evolution algorithm based on the constrained dropout neural network (CDNNEA) is proposed. The constrained dropout neural network is used as a scalable scheme to enhance the applicability of surrogate models in high-dimensional space. In the model management part
the convergence and diversity evaluation indicators of the population individuals are constructed
in order to adaptively select the representative individuals to guide the surrogate model to update. Experiments on DTLZ benchmark problems show that CDNNEA performs the best or approximates the best compared with other advanced algorithms. Meanwhile
CDNNEA is applied to the high-dimensional parameter adaptive optimization of practical problems. The experimental results show the proposed representative solutions have better recognition accuracy and controllable model computation in practical application
which verifies the effectiveness of CDNNEA in high-dimensional expensive optimization problems.
关键词
Keywords
references
KNOWLES J . ParEGO: A hybrid algorithm with on-line landscape approximation for expensive multiobjective optimization problems [J]. IEEE Transactions on Evolutionary Computation , 2006 , 10 ( 1 ): 50 - 66 .
ZHANG Q , LI H . MOEA/D: A multiobjective evolutionary algorithm based on decomposition [J]. IEEE Transactions on Evolutionary Computation , 2007 , 11 ( 6 ): 712 - 731 .
CHUGH T , JIN Y , MIETTINEN K , et al . A surrogate-assisted reference vector guided evolutionary algorithm for computationally expensive many-objective optimization [J]. IEEE Transactions on Evolutionary Computation , 2018 , 22 ( 1 ): 129 - 142 .
孙超利 , 李贞 , 等 . 模型辅助的计算费时进化高维多目标优化 [J]. 自动化学报 , 2022 , 48 ( 4 ): 1119 - 1128 .
SUN Chao-li , LI Zhen , et al . Surrogate-assisted expensive evolutionary many-objective optimization [J]. Acta Automatica Sinica , 2022 , 48 ( 4 ): 1119 - 1128 . (in Chinese)
PAN L , HE C , TIAN Y , et al . A classification-based surrogate-assisted evolutionary algorithm for expensive many-objective optimization [J]. IEEE Transactions on Evolutionary Computation , 2019 , 23 ( 1 ): 74 - 88 .
GUO D , WANG X , et al . Evolutionary optimization of high-dimensional multiobjective and many-objective expensive problems assisted by a dropout neural network [J]. IEEE Transactions on Systems, Man and Cybernetics , 2022 , 52 ( 4 ): 2084 - 2097 .
MCKAY M , BECKMAN R , CONOVER W . A comparison of three methods for selecting values of input variables in the analysis of output from a computer code [J]. Technometrics , 2000 , 42 ( 1 ): 55 - 61 .
SUN W , HU X , ZHANG L . Moments of discounted aggregate claims with dependence based on Spearman copula [J]. Journal of Computational and Applied Mathematics , 2020 , 377 : 112889 .
LI L , YEN G , SAHOO A , et al . On the estimation of Pareto front and dimensional similarity in many-objective evolutionary algorithm [J]. Information Sciences , 2021 , 563 : 375 - 400 .
SLAPAK E , GAZDA J , BUGAR G , et al . HetNet spatial topology design using mini-batch k -means clustering [C]// Proceedings of the 2019 IEEE 15th International Conference on the Experience of Designing and Application of CAD Systems . Piscataway : IEEE , 2019 : 1 - 4 .
刘冰洁 , 毕晓君 . 一种基于角度信息的约束高维多目标进化算法 [J]. 电子学报 , 2021 , 49 ( 11 ): 2208 - 2216 .
LIU Bing-jie , BI Xiao-jun . A constrained many-objective evolutionary algorithm based on angle information [J]. Acta Electronica Sinica , 2021 , 49 ( 11 ): 2208 - 2216 . (in Chinese)
TIAN Y , CHENG R , ZHANG X , et al . PlatEMO: A MATLAB platform for evolutionary multi-objective optimization [J]. IEEE Computational Intelligence Magazines , 2017 , 12 ( 4 ): 73 - 87 .
DEB K , GOYAL M . A combined genetic adaptive search for engineering design [J]. Computer Science and Informatics , 1996 , 26 ( 4 ): 30 - 45 .
TIAN Y , CHENG R , ZHANG X , et al . An indicator-based multiobjective evolutionary algorithm with reference point adaptation for better versatility [J]. IEEE Transactions on Evolutionary Computation , 2018 , 22 ( 4 ): 609 - 622 .
DENG L . The MNIST database of handwritten digit images for machine learning research [Best of the Web] [J]. IEEE Signal Processing Magazine , 2012 , 29 ( 6 ): 141 - 142 .
0
浏览量
10
下载量
1
CSCD
- 网络PDF下载
- Meta-XML下载
- BibTeX下载
- Endnote下载
- RefMan 下载
- NoteExpress下载
- NoteFirst下载
关联资源
相关文章
相关作者
相关机构
- 技术支持由北京北大方正电子有限公司提供 京ICP备09064830号-19
京公网安备11010802024621 - 本系统建议在Chrome、 IE9+ 以上版本浏览器阅读本站内容，360浏览器请切换至极速模式
- Cookies帮助我们提供服务并提供个性化体验。使用本网站，即表示您同意我们使用Cookies