摘要
本文研究了一种自组织结构的神经网络模式分类器．它由一个多层网络和一个自组织机构组成．通过对训练样本的聚类分析，它能有效地构造和训练网络，并能在学习过程中自适应地调节网络结构．它具有收敛速度快，能避免局部极值点和应用方便等显著优点．计算机模拟实验证实了其优越性．
This article studies a multilayer network classifier with self-organizing network architecture (SOMLN). SOMLN consists of a multilayer network and a self-organizing mechanism. It can both design and train a network effectively, and can also adjust the network archriecture during the learning period. SOMLN has advantages of fast learning speed, avoidnce of local minima and convenience for using. Computer simulation results verify the above characters of SOMLN.
出处
《信号处理》
CSCD
北大核心
1994年第2期75-80,86,共7页
Journal of Signal Processing
关键词
自组织
神经网络
模式分类器
Self organizing, neural network, Pattern classifier,clustering, learning
