Study on property of hidden neurons in feed-forward neural network

被引：0

作者：

Sheng, Ji-De ^{[1
]}

Zhang, Yan-Xin ^{[1
]}

Chang, Sheng-Jiang ^{[1
]}

Chen, Shu ^{[1
]}

机构：

[1] Inst. of Modern Optics, Nankai Univ., Tianjin 300071, China

来源：

Guangdianzi Jiguang/Journal of Optoelectronics Laser | 2001年 / 12卷 / 06期

关键词：

Backpropagation - Learning algorithms - Pattern recognition;

D O I：

暂无

中图分类号：

学科分类号：

摘要：

The multi-target distortion invariance pattern recognition of three airplane models is studied by employing feedforward neural networks and backpropagation algorithm. With previous results for the same task by using cascaded neural networks and clustering encoding, the number of the hidden neurons and its effect on recognizing rate and weight training are studied. An efficient method for reducing the hidden neurons and simplifying the recognizing network structure is proposed.

引用

页码：620 / 622