École Nationale Supérieure de
Techniques Avancées

Abstract

The objective of this research project is focused on high dimensional linear regression. In this context, the least squares estimator penalized by the L1 norm of the coefficient, known as LASSO, is a commonly used method for estimating the unknown parameter of the regression. However, for problems involving a large number of explanatory variables, the computation of LASSO can be computationally expensive in terms of computation time and resources. Therefore, the Least Angle Regression (LARS) algorithm presents itself as an efficient alternative for LASSO estimation in this specific context. Hence, we will conduct an in-depth study of the existing LARS algorithm and propose a novel approach to present it while enhancing its performance.