We propose a new sparsity-smoothness penalty for high-dimensional generalized additive models. The combination of sparsity and smoothness is crucial for mathematical theory as well as performance for finite-sample data. We present a computationally efficient algorithm, with provable numerical convergence properties, for optimizing the penalized likelihood. Furthermore, we provide oracle results which yield asymptotic optimality of our estimator for high dimensional but sparse additive models. Finally, an adaptive version of our sparsity-smoothness penalized approach yields large additional performance gains.
Publié le : 2009-12-15
Classification:
Group lasso,
model selection,
nonparametric regression,
oracle inequality,
penalized likelihood,
sparsity,
62G08,
62F12,
62J07
@article{1256303527,
author = {Meier, Lukas and van de Geer, Sara and B\"uhlmann, Peter},
title = {High-dimensional additive modeling},
journal = {Ann. Statist.},
volume = {37},
number = {1},
year = {2009},
pages = { 3779-3821},
language = {en},
url = {http://dml.mathdoc.fr/item/1256303527}
}
Meier, Lukas; van de Geer, Sara; Bühlmann, Peter. High-dimensional additive modeling. Ann. Statist., Tome 37 (2009) no. 1, pp. 3779-3821. http://gdmltest.u-ga.fr/item/1256303527/