Cross-validation estimations of hyper-parameters of Gaussian processes with inequality constraints

Abstract : In many situations physical systems may be known to satisfy inequality constraints with respect to some or all input parameters. When building a surrogate model of this system (like in the framework of computer experiments 7), one should integrate such expert knowledge inside the emulator structure. We proposed a new methodology to incorporate both equality conditions and inequality constraints into a Gaussian process emulator such that all conditional simulations satisfy the inequality constraints in the whole domain 6. An estimator called mode (maximum a posteriori) is calculated and satisfies the inequality constraints. Herein we focus on the estimation of covariance hyper-parameters and cross validation methods 1. We prove that these methods are suited to inequality constraints. Applied to real data in two dimensions, the numerical results show that the Leave-One-Out mean square error criterion using the mode is more efficient than the usual (unconstrained) Kriging mean.
Type de document :
Communication dans un congrès
Spatial Statistics 2015: Emerging Patterns committee, Jun 2015, Avignon, France. Procedia Environmental Sciences, 27, pp.38-44, 〈10.1016/j.proenv.2015.07.105〉
Liste complète des métadonnées

Littérature citée [9 références]  Voir  Masquer  Télécharger

https://hal.archives-ouvertes.fr/hal-01539509
Contributeur : Hassan Maatouk <>
Soumis le : jeudi 15 juin 2017 - 02:00:48
Dernière modification le : mardi 23 octobre 2018 - 14:36:08
Document(s) archivé(s) le : mardi 12 décembre 2017 - 12:32:56

Fichier

Maatouk_SS15.pdf
Fichiers éditeurs autorisés sur une archive ouverte

Identifiants

Collections

Citation

Hassan Maatouk, Olivier Roustant, Yann Richet. Cross-validation estimations of hyper-parameters of Gaussian processes with inequality constraints. Spatial Statistics 2015: Emerging Patterns committee, Jun 2015, Avignon, France. Procedia Environmental Sciences, 27, pp.38-44, 〈10.1016/j.proenv.2015.07.105〉. 〈hal-01539509〉

Partager

Métriques

Consultations de la notice

174

Téléchargements de fichiers

140