摘要
为了提高连续数值优化算法的普适性和鲁棒性,提出了基于自适应学习群体搜索技术的集成进化算法.该算法集成了3种自适应学习群体智能优化算法作为子算法,其中1种子算法是本文设计的,另外两种子算法来自相关文献.相应地,整个进化种群被分成了3个子种群,在进化过程中,算法以并行的方式采用每种子算法独立地进化各自的子种群,而在进化过程的不同阶段,每种子算法的进化策略及其参数可以自适应地调整.在实验部分,首先定义了算法性能度量标准,然后在26个较新的测试函数上做了算法性能对比实验,实验结果表明所提出的算法具有较高的普适性和鲁棒性.
In order to enhance the performance of universality and robustness of the numerical optimization algorithms, an ensemble of evolution algorithm based on self-adaptive learning population search techniques （EEA-SLPS） is proposed. EEA-SLPS integrates three self-adaptive learning population based algorithms from different fields of stochastic search techniques. One sub-algorithm is designed in this paper, the other two are recently proposed in relevant literature. The whole individual population is divided into three sub-populations, and each sub-algorithm is employed to evolve each sub-population respectively in parallel manner during the whole search process. In each sub-algorithm, both search strategies and parameters are gradually self-adaptive in different stages of the search process. The performance of EEA-SLPS is cxtensively evaluated on a suite of 26 bound-constrained test functions with different characteristics. By comparing with several state-of：the-art algorithms, the experimental results clearly verify the advantages of EEA-SLPS.
出处
《系统工程理论与实践》
EI
CSSCI
CSCD
北大核心
2014年第2期458-465,共8页
Systems Engineering-Theory & Practice
基金
江苏省普通高校研究生科研创新计划(CXLX11_0203)
国防基础研究基金(Q072006C002_1)
航空科学基金(2010zc13012)
关键词
自适应
集成进化
进化学习
智能计算
优化
self-adaptation
ensemble evolution
evolution learning
computational intelligence
optimization
