东北大学学报（自然科学版）

东北大学学报(自然科学版) ›› 2011, Vol. 32 ›› Issue (4): 472-475+480.DOI: -

• 论著 • 上一篇    下一篇

基于残差自然幂法的增量线性判别分析方法

陈东岳;吴成东;   

  1. 东北大学信息科学与工程学院;
  • 收稿日期:2013-06-19 修回日期:2013-06-19 发布日期:2013-04-04
  • 通讯作者: -
  • 作者简介:-
  • 基金资助:
    国家自然科学基金资助项目(61005032);;

Incremental linear discriminate analysis based on a residual natural power method

Chen, Dong-Yue (1); Wu, Cheng-Dong (1)   

  1. (1) School of Information Science and Engineering, Northeastern University, Shenyang 110819, China
  • Received:2013-06-19 Revised:2013-06-19 Published:2013-04-04
  • Contact: Chen, D.-Y.
  • About author:-
  • Supported by:
    -

RichHTML

0

PDF (PC)

789

摘要： 提出了将增量线性判别分析问题(LDA)转化为两个增量主元分析(PCA)问题的算法框架.为加速算法的收敛速度,推导了增量LDA中训练样本的类内离散度矩阵和协方差矩阵的无损实时更新公式,并在此基础上提出了一种基于残差协方差矩阵的自然幂增量PCA算法.将该增量PCA方法与基于双PCA结构的增量LDA算法框架相结合,实现了数据流的实时LDA处理.仿真结果表明,与已有的增量LDA方法相比,该方法在收敛速度、计算复杂度和可操作性上具有更优的性能.

关键词: 线性判别分析(LDA), 主元分析(PCA), 自然幂法, 无损更新, 增量算法

Abstract: A new algorithm with a structure containing two incremental principal component analysis (PCA) modules is proposed to solve the incremental linear discriminate analysis (LDA) problem. A residual covariance natural power (RCNP) PCA method and lossless update equations of both a within-class scatter matrix and a covariance matrix are also proposed for accelerating convergence of the incremental LDA. Simulation results show that the proposed method provides faster convergence, and better performance in complex computations and in ease of operation compared to other incremental LDAs.

中图分类号: