噪声环境下精英克隆选择算法的收敛性分析  

Convergence analysis of elitist clonal selection algorithm in noisy environment

作　　者：洪露[1] 龚成龙[1] 王经卓[1] 纪志成[2] 

机构地区：[1]淮海工学院电子工程学院,江苏连云港222005 [2]江南大学物联网工程学院,江苏无锡214122

出　　处：《控制理论与应用》2013年第11期1457-1461,共5页Control Theory & Applications

基　　金：国家自然科学基金资助项目(61174013);江苏高校优势学科建设工程资助项目

摘　　要：利用随机过程相关理论对加性噪声环境下精英策略克隆选择算法(ECSA)的全局收敛性进行了研究.首先采用有序对的状态表示方法构造精英克隆选择算法在噪声环境中的Markov链;然后将算法种群中最佳亲和度函数的进化过程转化为下鞅,利用鞅理论证明了种群最佳亲和度函数的全局收敛性;最后通过分析加性噪声环境下精英克隆选择算法的状态转移概率的特性,证明了精英克隆选择算法在加性噪声环境下最终能以概率1收敛到全局最优解.The global convergence of elitist clonal selection algorithm （ECSA） in additive noisy environment is studied by using stochastic process. The Markov chain that models ECSA in noisy environment is constructed by a new state representation method named ordered pair. The evolutionary sequence best fitness functions is described as a submartingale, and the global convergence of the best fitness functions is derived based on martingale theory. The transition probability characteristics of the clonal selection algorithm in additive noisy environment are analyzed; by this property we conclude that ECSA is eventually globally convergent with probability 1 in additive noisy environment.

关 键 词：克隆选择算法 精英策略 加性噪声 鞅理论 转移概率 

分 类 号：TP18[自动化与计算机技术—控制理论与控制工程]