一种基于和谐理论神经网络的分类算法  

A Sorting Algorithm Based on Harmony Theory Neural Network

作　　者：张冰[1] 朱明程[1] 

机构地区：[1]深圳大学信息工程学院,广东深圳518060

出　　处：《计算机应用》2004年第2期47-49,共3页journal of Computer Applications

基　　金：国家自然科学基金资助项目 (6 9976 0 2 0 )

摘　　要：文中从和谐理论神经网络派生一种新的神经网络模型,并以该模型为基础提出了一种新的并行分类算法。神经网络采用不完全连接方式,网络节点的连接数大大减少,具有计算功能简单便于硬件实现的优点。同时,文中提出的分类算法不同于大多数基于神经网络的分类算法,不需事先对待分类的数据进行任何形式的预处理,分类运算可在有限步长内自动完成,具有结构简单、透明度高、运算速度快等特点。This paper presents a parallel sorting algorithm based on neural network derived from the harmony theory neural network. The nodes of the proposed neural network are connected in an incomplete way, and the connections between the nodes are greatly reduced. It has the advantage of simple in function and easy for hardware implementation. At the same time, the proposed sorting algorithm, different from other artificial neural network based sorting schemes, does not need any pre-processing for the data to be sorted, and the sorting operation can be terminated automatically in a finite number of steps. The proposed sorting algorithm has the characteristics of simple in function and structure as well as transparent and rapid in operation.

关 键 词：并行分类算法 人工神经网络 和谐理论 

分 类 号：TP183[自动化与计算机技术—控制理论与控制工程]