A Composite Lasso Penalty With an Application in Cosmology

Jairo Diaz Rodriguez, Sylvain Sardy

Abstract: An image was taken by the XMM telescope to better understand galaxy clusters. We model the data as a high dimensional Poisson inverse problem and develop a lasso type estimator with a composite penalty (with two thresholds) to handle both estimation of the gas density in the cluster and detection of point sources.