English

Estimating composite functions by model selection

Statistics Theory 2013-01-29 v2 Statistics Theory

Abstract

We consider the problem of estimating a function ss on [1,1]k[-1,1]^{k} for large values of kk by looking for some best approximation by composite functions of the form gug\circ u. Our solution is based on model selection and leads to a very general approach to solve this problem with respect to many different types of functions g,ug,u and statistical frameworks. In particular, we handle the problems of approximating ss by additive functions, single and multiple index models, neural networks, mixtures of Gaussian densities (when ss is a density) among other examples. We also investigate the situation where s=gus=g\circ u for functions gg and uu belonging to possibly anisotropic smoothness classes. In this case, our approach leads to a completely adaptive estimator with respect to the regularity of ss.

Keywords

Cite

@article{arxiv.1102.2818,
  title  = {Estimating composite functions by model selection},
  author = {Yannick Baraud and Lucien Birgé},
  journal= {arXiv preprint arXiv:1102.2818},
  year   = {2013}
}

Comments

37 pages

R2 v1 2026-06-21T17:25:59.076Z