生存模糊自适应的蚁群算法及收敛性  

Ant colony algorithm with fuzzy adaptive survival and its convergence

作　　者：薛晗[1] 李迅[1] 马宏绪[1] 

机构地区：[1]国防科技大学机电工程与自动化学院,长沙410073

出　　处：《控制与决策》2009年第9期1288-1293,共6页Control and Decision

基　　金：国家自然科学基金项目(60475035);国家863计划项目(2007AA041603);湖南省科学技术厅重大科技专项计划项目(2007FJ1806)

摘　　要：首次将多元生存分析引入进化算法,设计了一种生存自适应的蚁群算法.对蚁群算法收敛过程建立生存模型,用Kaplan-Meier法计算生存时间估计值和生存函数曲线,以算法各参数作为生存时间的协变量,用COX比例危险率回归模型来定量分析其相互依赖关系,分析了种群大小对早熟收敛的影响.根据个体适应度和种群多样性对剩余生存时间进行模糊控制,实现种群规模的自适应调控.数值实例验证了算法的有效性、稳定性及准确性.Multivariate survival analysis is introduced into evolutionary algorithm for the first time. And an ant colony optimization algorithm with fuzzy adaptive survival time（FASTACO） is proposed. Parametric survival model with concomitant variables is built for the convergence process of ACO algorithm. Kaplan-Meier method is used to compute the estimated survival time and survival function curve. Parameters of the algorithm are regarded as the concomitant variables of survival time, and COX regression model is used to compute their dependence relationship. The influence of population size to the premature convergence is analyzed. The increment of remaining life time is automatically tuned by a fuzzy controller according to the fitness of ant and population variety, which realizes population size adaptation in the evolutionary process. A numerical example shows the effectiveness, stability and accuracy of the proposed algorithm.

关 键 词：蚁群算法 生存分析 模糊自适应 COX回归模型 KAPLAN-MEIER法 

分 类 号：TP387[自动化与计算机技术—计算机系统结构] O221[自动化与计算机技术—计算机科学与技术]