Feature selection techniques for spatial point processes intensity estimation by means of LASSO methods

Date(s) : 15/01/2018   iCal
15 h 30 min - 16 h 30 min

Many methods for estimating parametrically the intensity function of inhomogeneous spatial point processes are available in the literature. Almost all papers consider the setting where the number of covariates is moderate. Since more and more applications involve a large number of covariates, the existing methods become undesirable: all covariates are selected yielding an increasing standard error for parameter estimates. In my talk, I will present feature selection procedures based on LASSO-type techniques to deal with such data.

After an introduction and a research motivation, I will talk about standard parametric methods for spatial point process intensity estimation and their computational implementation. I will explain how our methods are developed and related to the existing methods, followed by asymptotic properties and computational implementation. Numerical results from a simulation study and an application will be presented to support our theoretical results. We discuss how these works could potentially be extended.

*Based on joint-work with Jean-François Coeurjolly and Frédérique Letué.


Catégories Pas de Catégories

Retour en haut