摘要
采用高斯变异算子的进化规划算法存在早熟现象,根本原因是高斯变异产生的变异量较小,导致个体分量乃至整个个体不发生变异.文中从变异算子、个体分量值的计算和搜索空间三个方面改进了进化规划算法.设计了能产生较大变异量的离散余弦变换算子,并且采用动态比例变异法动态调整个体中的每个分量,多个体竞争策略扩大了算法的搜索空间.针对复杂采购业务模型,运用改进的进化规划算法求解.实验证明,改进的算法在求解精度上优于采用高斯变异和随机变异的进化规划算法,解决了进化规划算法的早熟问题.
Evolutionary programming with Gauss mutation operator has premature convergence. The root causes is mutation value produced by Gauss mutation is so smaller that every variable in individual and individual itself may not be mutation. This paper improves evolutionary programming in three aspects of mutation operator, individual opponent value and search space. At first, Gauss mutation operator is replaced with the improved discrete cosine transformation operator which can produce more big mutation value; the formula of dynamic and proportional mutation designed can adjust every component value dynamically in individual, and multi-individual competition strategy enlarges its search space. Fhrthermore, an ordering model is created of complicate ordering business, and it is solved by using algorithm of improved evolutionary programming in Matlab. The simulated experiment result indicates that this improved algorithm is superior to evolutionary programming with Gauss mutation operator and random mutation operator in solution precision. This research has solved problem of premature convergence in standard evolutionary programming.
出处
《系统工程理论与实践》
EI
CSCD
北大核心
2009年第6期172-177,共6页
Systems Engineering-Theory & Practice
基金
湖北省自然科学基金(2007ABA003)
湖北省教育厅科学技术研究重点项目(D200711006)
关键词
进化规划
离散余弦变换
动态比例变异
采购方案优化
evolutionary programming
discrete cosine transform
dynamic proportional mutation
orderingplan optimization
