Asymptotic Distribution of a Single Minor Component Estimator

Abstract

The ordinary differential equation(ODE)method is applied to the adaptive algorithms, that is LMS type PAST algorithm and Krasulina/Reddy(KR) algorithm for extracting single minor component. The asymptotic distributions of these algorithms are derived. In computer simulations, the estimated variances of the eigenvector in these methods are compared with the theoretical values. We show that the LMS type PAST algorithm has a better convergence property than the KR algorithm.