一种改进型克隆选择算法及其几乎处处强收敛性研究  

On an improved clonal selection algorithm and its almost sure strong convergence

作　　者：洪露[1,2] 纪志成[2] 龚成龙[1] 

机构地区：[1]淮海工学院电子工程学院,江苏连云港222005 [2]江南大学通信与控制工程学院,江苏无锡214122

出　　处：《控制与决策》2010年第5期725-729,共5页Control and Decision

基　　金：中国博士后科学基金项目(200904501048);江苏省博士后基金项目(0901076C)

摘　　要：基于克隆选择原理,引入混沌机制和小生境技术,提出一种改进型克隆选择算法(ICSA).该算法比传统的克隆选择算法具有更好的种群多样性和全局寻优能力.以随机过程理论为数学工具,分析了ICSA所形成抗体种群的平均适应度函数的鞅性质,并由此得出算法几乎处处强收敛性的结论.进而证明了,当状态空间有限时,该算法能在有限步内以概率1收敛到全局最优.仿真实验表明,该算法能有效地抑制早熟,具有更好的全局收敛性.By integrating chaos mechanism and niche technique,an improved clonal selection algorithm（ICSA） is proposed based on the clonal selection principle.The algorithm not only maintains better population diversity than the classical clonal selection algorithm,but also converges to the global optimal solution rapidly.By using stochastic processes martingale theory,the martingale characteristic of the average fitness of the population is analyzed and the almost sure strong convergence of ICSA is deduced.Furthermore,it is proved that the algorithm is globally convergent with probability 1 in a finite number of steps when the state space is finite.The simulation results show that ICSA can inhibit prematurity and has preferable global convergent performance.

关 键 词：克隆选择算法 混沌 鞅 几乎处处强收敛 

分 类 号：TP18[自动化与计算机技术—控制理论与控制工程]