El Lasso: regularització i selecció de predictors

Abstract

[en] Lasso (Least Absolute Shrinkage and Selection Operator) is a regression method that performs both regularization and variable selection, improving prediction accuracy and interpretability of the resulting model. In this project we follow evolution from the plain linear model, through Ridge regression, for many years the most popular technique to improve the precision of predictions, to Lasso. We delve into numerical procedures for calculating Lasso solutions: coordinate descent and LARS. We see some extensions of Lasso such as Elastic Net regression, a neat improvement when optimality fails. We illustrate these methods with several real data examples using the R programming language (see notebooks and HTML files in appendices to the main text).