摘要
目前，广泛应用于解决模式识别问题的方法有：分类树和层次前馈神经网络．本文提出了一种基于神经树结构的模式分类新方法．该方法使用小规模的神经网络作为分类树的节点，提取模式中非线性特征信息．实验结果表明，该方法一方面可以减小分类树用于模式识别产生的误差和分类村中节点的数目，另一方面可以缩短训练神经网络所需的时间。
Classification trees and multilayer feed forward neural networks are two popular approaches to the pattern recognition problem. In this paper,a new pattern classification way based on neural treeis presented. This approach uses some small scale neural networks as the nodes of the classification tree to extract the nonlinear features of patterns. The experiment results have shown that this approach can decrease the error rate of pattern recognition and the number of nodes of classification tree,also can shorter the training time of neural network.
出处
《电子学报》
EI
CAS
CSCD
北大核心
1997年第7期107-110,共4页
Acta Electronica Sinica
基金
国家自然科学基金
关键词
神经网络
模式识别
特征提取
分类树
Neural network,Pattern recognition,Feature extraction,Classification tree
