摘要
遗传算法的一个具有代表性的应用领域就是发现一个已知的,通常比较复杂的系统的输入—输出映射。神经网络聚类方法中,比较著名的方法之一就是“竞争学习”。竞争学习采用若干单元的层次结构,以一种“胜利者全取”的方式对系统当前处理的对象进行竞争。通常的神经网络聚类方法,由于其较长的处理时间和数据的复杂性,很难适用于大型数据库。为此,文章采用遗传算法发现一个已知的、通常比较复杂的系统的输入—输出映射,利用调制的小波基对输入模式预处理,在函数链神经网络的基础上设计了一种基于遗传算法和小波变换函数链神经网络的竞争学习系统,充分利用遗传算法、小波变换和函数链神经网络的优势,这样设计的系统有惊人的学习速度、体系结构的通用性好、适应性强等特点,以此作为数据聚类分析工具,能够达到简化数据聚类的复杂性、缩短系统处理时间等效果。
An application with representativeness of the Genetic Algorithms finds one that has already known,usually relatively complicated systematic introduction exported and shines upon at.In the method of network cluster of nerve,one of the relatively famous methods is that ″competitive learning″.competitive learning adopts the hierarchical structures of units,carry on the competition by one a kind of way target that punish at present system on″ victor draw completely″s.The common nerve method of network cluster,because of its relatively long treatment time and complexity of the data,it is very difficult to suitable for the large-scale data base.For this reason,this literary grace finds one a piece of know alreadies with hereditary algorithms,usually relatively complicated systematic introduction exports and shines upon at-,utilize the little wave base modulated to inputting the mode pretreatment,Design each at hereditary algorithm and little wave study the system by type function chain nerve the competitions of networks on the basis of functional-link neural network,fully utilize the advantage of hereditary algorithm,small wave varying and function chain nerve network,system that design like this have surprising study speed,system the commonability characteristic fine,better adaptabilities of structure,regard this as the analysis tool of cluster of data,can reach the complexity of simplifying the cluster of the data and shorten and deal with the results,such as time,etc.systematically.
出处
《计算机工程与应用》
CSCD
北大核心
2005年第20期186-188,218,共4页
Computer Engineering and Applications
关键词
遗传算法
小波变换
函数链神经网络
竞争学习
Genetic Algorithms,wavelets transforms,functional-link neural network,competitive learning
