An effective offspring generation strategy for many-objective optimization driven by knee points under variable classification

doi:10.1007/s10489-022-03307-8

	An effective offspring generation strategy for many-objective optimization driven by knee points under variable classification
	Wei, Li-sen; Li, Er-chao
刊名	APPLIED INTELLIGENCE
	2022
关键词	Effective solution generation strategy Variable classification Parallel mutation Many-objective optimization
ISSN号	0924-669X
DOI	10.1007/s10489-022-03307-8
英文摘要	In many-objective optimization problems, the difficulty of optimization increases as the number of targets increases. When the number of objectives increases, the individuals become extremely sparse in the objective space and the performance of environmental selection strategies weaken. To solve this problem, we proposed an effective offspring generation strategy driven by knee points under variable classification, termed VKOS. In VKOS, there is no mating selection, and the excellent genes of the knee points are used to guide the generation of outstanding individuals with different attributes directionally. Specifically, first, the convergence variables and diversity variables of the problem are obtained by variable classification; then identify the knee points of the current population, and finally propose a parallel mutation method to mutate the current population to obtain offspring. In order to verify the versatility and effectiveness of the strategy, RPEA-VKOS, NSGAIII-VKOS, RPDNSGAII-VKOS, MaOEA/IBP-VKOS compared with the original algorithm on 16 widely used benchmark problems. In addition, taking RPEA as an example, The VKOS proposed in this paper compared with several single mutation operators DE, PLM, NUM and representative solution generation strategies DEMR, VCEM, MM on these benchmark problems. The extensive experiments on well-known benchmark problems ranging from 3 to 15 objectives show that VKOS improves the performance of MaOEAs and has the superior performance over three single mutation operators and typical offspring generation strategies when solving most of these test MaOPs.
WOS研究方向	Computer Science
语种	英语
出版者	SPRINGER
WOS记录号	WOS:000777414800004
内容类型	期刊论文
源URL	[http://ir.lut.edu.cn/handle/2XXMBERH/158085]
专题	电气工程与信息工程学院
作者单位	Lanzhou Univ Technol, Coll Elect Engn & Informat Engn, Lanzhou 730050, Gansu, Peoples R China
推荐引用方式 GB/T 7714	Wei, Li-sen,Li, Er-chao. An effective offspring generation strategy for many-objective optimization driven by knee points under variable classification[J]. APPLIED INTELLIGENCE,2022.
APA	Wei, Li-sen,&Li, Er-chao.(2022).An effective offspring generation strategy for many-objective optimization driven by knee points under variable classification.APPLIED INTELLIGENCE.
MLA	Wei, Li-sen,et al."An effective offspring generation strategy for many-objective optimization driven by knee points under variable classification".APPLIED INTELLIGENCE (2022).