High-dimensional linear regression and variable selection with the Lasso (Pascaline Descloux, University of Geneva, Switzerland)

16.04.2018 17:00 – 18:00

In modern regression problems it has become common that the number of parameters to estimate exceeds the number of observations. In such a high-dimensional setting, the parameters are often assumed to be sparse; in other words only a few of the potential predictors are relevant. In this talk I will define the Lasso (Least Absolute Shrinkage and Selection Operator), a sparse estimator which is commonly used in high-dimensional linear regression. I will then introduce some notions of convex optimization which will allow us to characterize the lasso solution and understand how it performs variable selection.

Lieu

Bâtiment: Centre Acacias

Room 17

entrée libre