信息与控制2017，Vol.46Issue(5)：536-542,7.DOI:10.13976/j.cnki.xk.2017.0536
一种耦合的主/次特征对提取神经网络算法
A Coupled Principal/Minor Eigen-pairs Extraction Neural Network Algorithm
摘要
Abstract
At present, most principal component analysis ( PCA) algorithms based on neural networks are consid-ered non-coupled algorithms, which suffer from the so-called speed-stability problem. In recent years, a few coupled algorithms have been proposed, and these algorithms can mitigate the speed-stability problem of most non-coupled algorithms. However, these coupled algorithms still suffer from problems such as a complex deri-vation process and highly complex expression. In this paper, we propose two coupled PCA algorithms based on a novel information criterion and by modifying Newton′s method, and we analyze the convergence of algo-rithms by using Jacobian matrix. Compared with existing coupled algorithms, the proposed algorithms can be obtained easily without the need to calculate the inverse Hessian matrix. Simulation experiments validate the favorable capability of the proposed algorithms.关键词
主成分分析/神经网络/"速度—稳定"问题/耦合算法/牛顿方法Key words
PCA( principal component analysis)/neural networks/speed-stability problem/coupled algorithm/Newton′s method分类
信息技术与安全科学引用本文复制引用
汪奔,孔祥玉,冯晓伟,曹泽豪..一种耦合的主/次特征对提取神经网络算法[J].信息与控制,2017,46(5):536-542,7.基金项目
国家自然科学基金资助项目(61374120,61673387) （61374120,61673387）
