Geodesic
On the use of sequential Monte Carlo methods for approximating smoothing functionals, with application to fixed parameter estimation
Jimmy Olsson1 ; Olivier Cappé2 ; Randal Douc3 ; Éric Moulines2
1 Center for Mathematical Sciences, Lund University, Sweden
2 Laboratoire Traitement et Communication de l'Information, CNRS / GET Télécom Paris, France
3 Centre de Mathématiques APpliquées, Ecole Polytechnique, France
ESAIM. Proceedings, Tome 19 (2007), pp. 6-11.

Voir la notice de l'article provenant de la source EDP Sciences

Sequential Monte Carlo (SMC) methods have demonstrated a strong potential for inference on the state variables in Bayesian dynamic models. In this context, it is also often needed to calibrate model parameters. To do so, we consider block maximum likelihood estimation based either on EM (Expectation-Maximization) or gradient methods. In this approach, the key ingredient is the computation of smoothed sum functionals of the hidden states, for a given value of the model parameters. It has been observed by several authors that using standard SMC methods for this smoothing task requires a substantial number of particles and may be unreliable for larger observation sample sizes. We introduce a simple variant of the basic sequential smoothing approach based on forgetting ideas. This modification, which is transparent in terms of computation time, reduces the variability of the approximation of the sum functional. Under suitable regularity assumptions, it is shown that this modification indeed allows a tighter control of the Lp error of the approximation.
DOI : 10.1051/proc:071902

Jimmy Olsson 1 ; Olivier Cappé 2 ; Randal Douc 3 ; Éric Moulines 2

1 Center for Mathematical Sciences, Lund University, Sweden
2 Laboratoire Traitement et Communication de l'Information, CNRS / GET Télécom Paris, France
3 Centre de Mathématiques APpliquées, Ecole Polytechnique, France 
@article{EP_2007_19_a2,
     author = {Jimmy Olsson and Olivier Capp\'e and Randal Douc and \'Eric Moulines},
     title = {On the use of sequential {Monte} {Carlo} methods for approximating smoothing functionals, with application to fixed parameter estimation},
     journal = {ESAIM. Proceedings},
     pages = {6--11},
     publisher = {mathdoc},
     volume = {19},
     year = {2007},
     doi = {10.1051/proc:071902},
     language = {en},
     url = {http://geodesic.mathdoc.fr/articles/10.1051/proc:071902/}
}
TY  - JOUR
AU  - Jimmy Olsson
AU  - Olivier Cappé
AU  - Randal Douc
AU  - Éric Moulines
TI  - On the use of sequential Monte Carlo methods for approximating smoothing functionals, with application to fixed parameter estimation
JO  - ESAIM. Proceedings
PY  - 2007
SP  - 6
EP  - 11
VL  - 19
PB  - mathdoc
UR  - http://geodesic.mathdoc.fr/articles/10.1051/proc:071902/
DO  - 10.1051/proc:071902
LA  - en
ID  - EP_2007_19_a2
ER  - 
%0 Journal Article
%A Jimmy Olsson
%A Olivier Cappé
%A Randal Douc
%A Éric Moulines
%T On the use of sequential Monte Carlo methods for approximating smoothing functionals, with application to fixed parameter estimation
%J ESAIM. Proceedings
%D 2007
%P 6-11
%V 19
%I mathdoc
%U http://geodesic.mathdoc.fr/articles/10.1051/proc:071902/
%R 10.1051/proc:071902
%G en
%F EP_2007_19_a2
Jimmy Olsson; Olivier Cappé; Randal Douc; Éric Moulines. On the use of sequential Monte Carlo methods for approximating smoothing functionals, with application to fixed parameter estimation. ESAIM. Proceedings, Tome 19 (2007), pp. 6-11. doi : 10.1051/proc:071902. http://geodesic.mathdoc.fr/articles/10.1051/proc:071902/

Cité par Sources :