Fast and stable coupled minor component analysis rules

Abstract

Coupled learning algorithm, in which the eigenvector and eigenvalue of a covariance matrix are estimated in coupled equations simultaneously, is a solution to the speed-stability problem that plagues most noncoupled learning rules. Moller has proposed a class of well-performed CPCA (coupled principal component analysis) algorithms, but it is a pity that only few of CMCA (coupled minor component analysis) algorithm was proposed until now. In this paper, to expand the CMCA field, we propose some stable CMCA algorithms based on Moller's CPCA and CMCA algorithms. The proposed algorithms provide efficient methods to extract the minor eigenvector and eigenvalue of a covariance matrix. Simulation experiments confirm the effectiveness of the proposed algorithms.