作　　者： ; ; ;

机构地区： 西北工业大学机电学院系统集成与工程管理研究所

出　　处： 《计算机工程与应用》 2004年第23期1-3,53,共4页

摘　　要： 由于免疫系统在特征抽取、模式识别、学习、记忆等方面强大的信息处理功能以及其自身的分布式特性,所以免疫算法出现,并逐渐应用于许多工程实际。然而,现有的免疫算法模型存在不少缺陷,缺乏有效的收敛特性。问题主要出在抗体的评价、促进和抑制以及记忆库的使用上。该文提出了基于动态评价的免疫算法,在算法中引入临时解集,对抗体进行多次、动态的评价、促进和抑制,这大大提高了算法的收敛速度,并有效防止算法陷入局部最优。另外,记忆库自始至终一直被利用,有利于算法快速收敛。最后,用马尔可夫链描述了该算法,并证明了算法的收敛性。 Owning to immune system's powerful information processing capabilities,such as feature extraction,pattern recognition,learning,memory,and its distributive nature,immune algorithm(IA)has emerged,and gradually been applied to many engineering practices.However,there are limitations such as the lack of convergence efficiency in the existing immune algorithm models.This problem is caused by the evaluation form,the proliferation and suppression of antibodies,and the use of the memory library.Thus the paper improves the existing immune algorithm,and then proposes dynamic evaluation based immune algorithm.By introducing the temporary solution group,the evaluation,proliferation and suppression of antibodies take more times in a dynamic way,which updates the solution group and improves the convergence speed of the algorithm in later calculation.Besides,the memory library is used from the beginning to the end,which also helps to accelerate the convergence speed.Last,the proposed algorithm is described with Markov chain,and the convergence of the algorithm is proved.

关 键 词： 免疫系统 免疫算法 动态评价 临时解集 抗体 记忆库 马尔可夫链 收敛性

领　　域： [自动化与计算机技术] [自动化与计算机技术]