摘要
将正交试验设计引入到克隆选择操作中,设计出基于正交试验的克隆选择操作(clonal selection operation based on orthogonal experiment design,简称CSO-OED),并将其加入到典型的克隆选择算法中,设计出并联式的CSO+CSO-OED(Ⅰ)算法和串联式的CSO+CSO-OED(Ⅱ)算法.将新设计的算法用于9个经典的测试函数和6个复杂的测试函数进行对比测试,实验结果表明,CSO-OED能够有效地保持种群的多样性,避免算法不成熟收敛.CSO+CSO-OED(Ⅰ)和CSO+CSO-OED(Ⅱ)将全局搜索和局部搜索分开进行优化,对比实验表明,这种搜索策略不但能够保证算法的收敛性,还能有效地提高搜索解的精度,增强算法的鲁棒性.
This paper presents a clonal selection operation: clonal selection operation based on orthogonal experiment design (CSO-OED). This design is later combined with the typical clonal selection operation and results in two algorithms: CSO+CSO-OED(Ⅰ) adopting parallel mechanism and CSO+ CSO-OED(Ⅱ) adopting series mechanism. The validation in 9 classical benchmark functions and 6 complex functions has showed that CSO-OED can not only maintain the diversity of population, but also help avoid premature. Implemented in CSO+CSO-OED(Ⅰ) and CSO+CSO-OED(Ⅱ), the strategy that separates the local search and global search can not only guarantee the convergence but also improve the accuracy of global solution and the robustness of the algorithm.
出处
《软件学报》
EI
CSCD
北大核心
2010年第5期950-967,共18页
Journal of Software
基金
国家自然科学基金(Nos.60703107
60703108)
国家高技术研究发展计划(863)No.2009AA12Z210
国家重点基础研究发展计划(973)No.2006CB705700
新世纪优秀人才支持计划No.NCET-08-0811~~
关键词
人工智能
进化算法
人工免疫
克隆选择算法
正交试验设计
函数优化
artificial immune
evolutionary algorithm
artificial immune
clonal selection algorithm
orthogonal experiment design
function optimization