Title :
A Regularized Estimator For Linear Regression Model With Possibly Singular Covariance
Author :
Hoang, Hong Son ; Baraille, Rémy
Author_Institution :
SHOM/HOM, Toulouse, France
Abstract :
A regularized estimator is proposed for regression models in the case where the covariances may be singular. Conditions guaranteeing proximity of a regularized estimator to the optimal estimator are obtained by appropriate choice of regularization parameters by allowing a prescribed level of uncertainty. A simple Monte-Carlo simulation study is reported to highlight some aspects and performance of the proposed approach.
Keywords :
Monte Carlo methods; covariance matrices; regression analysis; Monte-Carlo simulation; linear regression model; optimal estimator; regularized estimator; singular covariance; Convergence; Eigenvalues and eigenfunctions; Estimation error; Mathematical model; Stability analysis; Vectors; Covariance matrix; LMS algorithm; linear regression system; parameter estimation; regularization;
Journal_Title :
Automatic Control, IEEE Transactions on
DOI :
10.1109/TAC.2012.2203552
