Computer Engineering and Applications ›› 2012, Vol. 48 ›› Issue (26): 22-24.

Previous Articles     Next Articles

MCA learning algorithm without restriction of initial weight vector

LI Xiaobo, FAN Yangyu   

  1. School of Electronics & Information, Northwestern Polytechnical University, Xi’an 710129, China
  • Online:2012-09-11 Published:2012-09-21

一种不受初始权值向量影响的MCA学习算法

李晓波,樊养余   

  1. 西北工业大学 电子信息学院,西安 710129

Abstract: Adaptive total least square algorithms based on MCA are not convergent when initial weight vector is not appropriate. A new MCA algorithm without restriction of initial weight vector is proposed. The convergence condition and domain of the proposed MCA learning algorithm are derived. Simulation results indicate that the proposed algorithm is effective in obtaining total least square solution.

Key words: total least square, minor component analysis, initial weight vector

摘要: 最小主元分析(Minor Component Analysis,MCA)类自适应总体最小二乘算法易受初始权值向量的影响而无法收敛。为解决这一问题,提出了一种不受初始权值向量影响的MCA学习算法,推导出了该算法的收敛条件与最终收敛域,并通过计算机仿真验证了该算法的正确性。

关键词: 总体最小二乘, 最小主元分析, 初始权值向量