摘要
自组织映射(SOM)神经网络存在训练结果不稳定、识别率低的问题。为此,提出一种斥力平衡SOM神经网络方法(RESOM)。该方法使用斥力原理,根据SOM模式区域在斥力场中迁移到的斥力平衡位置进行模式识别。比较普通SOM神经网络、SVM、RESOM对故障模式的识别结果表明,该方法训练时间少、识别正确率高、识别结果稳定。
For the problems of the instable training results and low rate of pattern recognition in Self-Organizing Mapping（SOM）,a novel Repulsive Equilibrium SOM（RESOM） is proposed in this paper.Theory of repulsive force is used in this method,pattern recognition based on the cluster regions of SOM,these regions move to repulsive equilibrium position in repulsion field.Experimental results show that RESOM method outperforms the normal SOM and SVM.It can improve stability and recognition rate of pattern recognition and reduce training time
出处
《计算机工程》
CAS
CSCD
北大核心
2011年第13期199-201,共3页
Computer Engineering
基金
重庆万州区科委基金资助项目(20084033)
重庆三峡学院基金资助项目(2008-sxxyqn-027)
关键词
自组织映射
斥力平衡
可信故障诊断
聚类迁移
神经网络
Self-Organizing Mapping（SOM）
repulsive equilibrium
trusted fault diagnosis
clustering migration
neural net
