Adaptive sequential design for regression on Schauder Basis

Serge Cohen (1) , Sébastien Gadat (1)

1 IMT - Institut de Mathématiques de Toulouse UMR5219

We present a new sequential algorithm to build both optimal design and model selection in a multi-resolution family of functions. This algorithm relies on a localization property of discrete sequential D and A-optimal designs for Schauder Basis. We use these property with a simulated annealing strategy to obtain our stochastic algorithm. We illustrate its efficiency on several numerical experiments.

Mots clés: Sequential Optimal Design; Model Selection; Stochastic Algorithm

Domaines: Statistiques [math.ST]; Théorie [stat.TH]

Serge Cohen, Sébastien Gadat. Adaptive sequential design for regression on Schauder Basis. 2008. ⟨hal-00358727⟩