Multidimensional Laplace formulas for nonlinear Bayesian estimation

Paul Bui Quang 1, 2 Christian Musso 2 François Le Gland 1
1 ASPI - Applications of interacting particle systems to statistics
IRMAR - Institut de Recherche Mathématique de Rennes, Inria Rennes – Bretagne Atlantique
Abstract : The Laplace method and Monte Carlo methods are techniques to approximate integrals which are useful in nonlinear Bayesian computation. When the model is one-dimensional, Laplace formulas to compute posterior expectations and variances have been proposed by Tierney, Kass and Kadane (1989). We provide in this article formulas for the multidimensional case. We demonstrate the accuracy of these formulas and show how to use them in importance sampling to design an importance density function which reduces the Monte Carlo error.
Type de document :
Communication dans un congrès
Proceedings of the 20th European Signal Processing Conference, Bucharest 2012, Aug 2012, Bucharest, Romania. 2012, 〈http://ieeexplore.ieee.org/document/6334291/〉
Liste complète des métadonnées

https://hal.inria.fr/hal-00911786
Contributeur : Francois Le Gland <>
Soumis le : vendredi 29 novembre 2013 - 20:13:00
Dernière modification le : jeudi 15 novembre 2018 - 11:58:38

Identifiants

  • HAL Id : hal-00911786, version 1

Citation

Paul Bui Quang, Christian Musso, François Le Gland. Multidimensional Laplace formulas for nonlinear Bayesian estimation. Proceedings of the 20th European Signal Processing Conference, Bucharest 2012, Aug 2012, Bucharest, Romania. 2012, 〈http://ieeexplore.ieee.org/document/6334291/〉. 〈hal-00911786〉

Partager

Métriques

Consultations de la notice

505