GdT Stat Optim : Gluing properties of SLOPE in low dimension
par
Tomasz Skalski(Wroclaw University of Science and Technology / LAREMA, Université d’Angers)
→
Europe/Paris
salle René Baire (IMB)
salle René Baire
IMB
Description
We consider a low-dimensional linear regression problem. We focus on the necessary and sufficient conditions for gluing properties and the support recovery of a SLOPE estimator in the orthogonal design. Afterwards we extend the problem to the consistency of SLOPE in the asymptotic low-dimensional case after relaxing the orthogonality assumption. Later on we discuss the support recovery of the low-dimensional SLOPE.
The talk is based on a joint research with M. Bogdan, P. Graczyk, B. Kolodziejek, P. Tardivel and M. Wilczynski.