Nous présentons de nouveaux algorithmes Monte Carlo par chaînes de Markov en interaction pour la résolution de processus à valeurs mesures non linéaires à temps discret. Ces modèles appartiennent à la classe des chaînes de Markov en auto interaction avec leurs mesures d'occupations. Nous proposons une variété de résultats de convergence, avec notamment des estimations exponentielles et un théorème de convergence uniforme par rapport au paramètre temporel. Cette analyse semble être la premiere de ce type pour des chaînes de Markov en auto-interaction. Nous illustrons ces modèles dans le cadre de semigroupes de Feynman–Kac couramment utilisés en physique, en biologie, et en statistiques.
We present a new class of interacting Markov chain Monte Carlo methods to approximate numerically discrete-time nonlinear measure-valued equations. These stochastic processes belong to the class of self-interacting Markov chains with respect to their occupation measures. We provide several convergence results for these new methods including exponential estimates and a uniform convergence theorem with respect to the time parameter, yielding what seems to be the first results of this kind for this type of self-interacting models. We illustrate these models in the context of Feynman–Kac distribution semigroups arising in physics, biology and in statistics.
Accepté le :
Publié le :
@article{CRMATH_2010__348_1-2_79_0, author = {Del Moral, Pierre and Doucet, Arnaud}, title = {A new class of interacting {Markov} chain {Monte} {Carlo} methods}, journal = {Comptes Rendus. Math\'ematique}, pages = {79--83}, publisher = {Elsevier}, volume = {348}, number = {1-2}, year = {2010}, doi = {10.1016/j.crma.2009.11.006}, language = {en}, url = {http://www.numdam.org/articles/10.1016/j.crma.2009.11.006/} }
TY - JOUR AU - Del Moral, Pierre AU - Doucet, Arnaud TI - A new class of interacting Markov chain Monte Carlo methods JO - Comptes Rendus. Mathématique PY - 2010 SP - 79 EP - 83 VL - 348 IS - 1-2 PB - Elsevier UR - http://www.numdam.org/articles/10.1016/j.crma.2009.11.006/ DO - 10.1016/j.crma.2009.11.006 LA - en ID - CRMATH_2010__348_1-2_79_0 ER -
%0 Journal Article %A Del Moral, Pierre %A Doucet, Arnaud %T A new class of interacting Markov chain Monte Carlo methods %J Comptes Rendus. Mathématique %D 2010 %P 79-83 %V 348 %N 1-2 %I Elsevier %U http://www.numdam.org/articles/10.1016/j.crma.2009.11.006/ %R 10.1016/j.crma.2009.11.006 %G en %F CRMATH_2010__348_1-2_79_0
Del Moral, Pierre; Doucet, Arnaud. A new class of interacting Markov chain Monte Carlo methods. Comptes Rendus. Mathématique, Tome 348 (2010) no. 1-2, pp. 79-83. doi : 10.1016/j.crma.2009.11.006. http://www.numdam.org/articles/10.1016/j.crma.2009.11.006/
[1] Feynman–Kac Formulae. Genealogical and Interacting Particle Systems, with Applications, Probability and Its Applications, Springer-Verlag, New York, 2004
[2] On convergence of chains with time empirical self-interactions, Proc. Royal Soc. Lond. A, Volume 460 (2003), pp. 325-346
[3] P. Del Moral, A. Doucet, Interacting Markov chain Monte Carlo methods for solving nonlinear measure-valued equations, INRIA Research Report, January 2008, Ann. Appl. Probab. (2010), http://www.imstat.org/aap/future_papers.html, in press
[4] Sequential Monte Carlo for Bayesian computation (with discussion), Bayesian Statistics, vol. 8, Oxford University Press, 2007, pp. 1-34
[5] Sequential Monte Carlo Methods in Practice (Doucet, A.; de Freitas, N.; Gordon, N., eds.), Statistics for Engineering and Information Science, Springer-Verlag, New York, 2001
[6] Monte Carlo Methods for Scientific Computing, Statistics, Springer-Verlag, New York, 2001
[7] Monte Carlo Statistical Methods, Springer-Verlag, New York, 2004
Cité par Sources :