计算机工程与应用 ›› 2021, Vol. 57 ›› Issue (22): 110-124.DOI: 10.3778/j.issn.1002-8331.2011-0366

• 理论与研发 • 上一篇    下一篇

基于全组合策略的多目标阴阳对算法

李大海,艾志刚,王振东   

  1. 江西理工大学 信息工程学院,江西 赣州 341000
  • 出版日期:2021-11-15 发布日期:2021-11-16

Improved Yin-Yang-Pair Multi-objective Optimization Algorithm Based on Full-Combination Strategy

LI Dahai, AI Zhigang, WANG Zhendong   

  1. School of Information Engineering, Jiangxi University of Science & Technology, Ganzhou, Jiangxi 341000, China
  • Online:2021-11-15 Published:2021-11-16

摘要:

基于前沿的阴阳对优化算法(Front-based Yin-Yang-Pair Optimization,F-YYPO)是一种新颖的轻量级多目标优化算法,其利用两点——局部开发点[Pi1]和全局探索点[Pi2]在搜索过程中的迭代交换实现搜索。基于F-YYPO提出了一种改进的多目标优化算法F-ACYYPO。新算法对F-YYPO做了以下三方面的改进:(1)对多个目标函数进行全组合,以增强优化个体分布的均匀性;(2)引入已在YYPO算法中被证明有明显性能提高效果的缩放因子[α]自适应措施;(3)改进F-YYPO存档操作的更新方式。采用在2009年进化计算大会多目标优化算法竞赛中使用的UF测试套件以及PlatEMO平台下的DTLZ测试套件进行算法的性能评估,将F-ACYYPO与F-YYPO以及其他多种已知性能优良的多目标优化算法NSGA2、SPEA2、MOPSO、MOGWO、gamultiobj、MOEA\D、GDE3进行性能测试及比较,并通过两个综合性指标(反转世代距离IGD、超体积HV)和一个收敛性指标(世代距离GD)进行性能评价。实验结果表明,F-ACYYPO比F-YYPO具有更高的计算精度以及更快的收敛速度,并且与其他高性能多目标算法相比,F-ACYYPO表现出了很强的竞争性,在综合性能指标下有将近超1/2的测试用例占优。

关键词: 多目标优化, 基于前沿的阴阳对优化(F-YYPO), 阴阳对优化

Abstract:

The Front-based Yin-Yang-Pair Optimization(F-YYPO) algorithm is a novel lightweight multi-objective optimization algorithm that uses the exchange of two points of the global exploration point [Pi2] and the local development point [Pi1] iteratively to archive target searching. This paper proposes an improved multi-objective optimization algorithm F-ACYYPO based on F-YYPO. F-ACYYPO improves the original F-YYPO in the following three aspects:(1)Adopt the complete combination of multiple objective functions to increase the optimal distribution of optimization individuals;(2) Adopt the adaptive tuning of scaling factor α, which had been proven to be an effective improvement to the YYPO algorithm;(3)Improve the update mode of archive operation of F-YYPO. This paper adopts UF test suite that had been introduced in the multi-objective optimization algorithm competition in the 2009 Evolutionary Computing Conference and DTLZ test suite from the PlatEMO platform to evaluate the performance of the proposed new algorithm and several state-of-the-art multi-objective optimization algorithms including NSGA2, SPEA2, MOPSO, MOGWO, gamultiobj, MOEA\D and GDE3. The performance of all algorithms is evaluated by two comprehensive indexes, the Inverted Generational Distance(IGD) and the Hyper Volume(HV), and a convergence index, the Generational Distance(GD) is also applied. Experimental results show that F-ACYYPO can achieve higher computational accuracy and faster convergence speed than the original F-YYPO algorithm. Compared with other multi-objective optimization algorithms, F-ACYYPO shows highly competitive performance and can even dominate nearly more than 1/2 test cases under two comprehensive evaluation indexes.

Key words: multi-objective optimization, Front-based Yin-Yang-Pair Optimization(F-YYPO), Yin-Yang-Pair optimization