Initialement motivés par des problèmes numériques d’analyse en temps long d’une diffusion ergodique, nous abordons ici la question suivante : si une diffusion Brownienne ergodique a une unique probabilité invariante, quelles sont les probabilités invariantes associées à sa dupliquée, i.e. au système formé par deux copies de la diffusion initiale. Nous nous focalisons notamment sur le cas où ces deux copies sont dirigées par le même mouvement Brownien (-point motion). Sous cette hypothèse, nous montrons que l’unicité de la probabilité invariante relative à la diffusion dupliquée (confluence faible) est essentiellement toujours vraie en dimension . En dimension supérieure, après avoir exhibé un contre-exemple, nous proposons une série de critères de confluence faible (de type intégral) mais aussi de confluence trajectorielle presque sûre, lisibles sur les coefficients de la diffusion au travers d’un exposant de Lyapounov non-infinitésimal. Ces critères permettent en particulier de traiter des cas non triviaux comme certaines classes de systèmes-gradients à potentiel (sur-quadratique) non convexe ou, à l’inverse, des systèmes pour lesquels la confluence est induite par le coefficient de diffusion. Nous montrons enfin que la propriété de confluence faible est associée à un problème de transport optimal. Dans un second temps, nous appliquons nos résultats pour optimiser la mise en œuvre de l’extrapolation Richardson–Romberg pour l’approximation par simulation de mesures invariantes.
With a view to numerical applications we address the following question: given an ergodic Brownian diffusion with a unique invariant distribution, what are the invariant distributions of the duplicated system consisting of two trajectories? We mainly focus on the interesting case where the two trajectories are driven by the same Brownian path. Under this assumption, we first show that uniqueness of the invariant distribution (weak confluence) of the duplicated system is essentially always true in the one-dimensional case. In the multidimensional case, we begin by exhibiting explicit counter-examples. Then, we provide a series of weak confluence criterions (of integral type) and also of a.s. pathwise confluence, depending on the drift and diffusion coefficients through a non-infinitesimal Lyapunov exponent. As examples, we apply our criterions to some non-trivially confluent settings such as classes of gradient systems with non-convex potentials or diffusions where the confluence is generated by the diffusive component. We finally establish that the weak confluence property is connected with an optimal transport problem. As a main application, we apply our results to the optimization of the Richardson–Romberg extrapolation for the numerical approximation of the invariant measure of the initial ergodic Brownian diffusion.
@article{AIHPB_2015__51_4_1562_0,
author = {Lemaire, Vincent and Pag\`es, Gilles and Panloup, Fabien},
title = {Invariant measure of duplicated diffusions and application to Richardson--Romberg extrapolation},
journal = {Annales de l'I.H.P. Probabilit\'es et statistiques},
volume = {51},
year = {2015},
pages = {1562-1596},
doi = {10.1214/13-AIHP591},
mrnumber = {3414458},
language = {en},
url = {http://dml.mathdoc.fr/item/AIHPB_2015__51_4_1562_0}
}
Lemaire, Vincent; Pagès, Gilles; Panloup, Fabien. Invariant measure of duplicated diffusions and application to Richardson–Romberg extrapolation. Annales de l'I.H.P. Probabilités et statistiques, Tome 51 (2015) pp. 1562-1596. doi : 10.1214/13-AIHP591. http://gdmltest.u-ga.fr/item/AIHPB_2015__51_4_1562_0/
[1] and . Stability in distribution for a class of singular diffusions. Ann. Probab. 20 (1) (1992) 312–321. | MR 1143422 | Zbl 0749.60073
[2] . Statistical equilibrium and two-point motion for a stochastic flow of diffeomorphisms. In Spatial Stochastic Processes 189–218. Progress in Probability 19. Birkhäuser, Boston, MA, 1991. | MR 1144097 | Zbl 0744.60063
[3] and . Large deviations and stochastic flows of diffeomorphisms. Probab. Theory Related Fields 80 (2) (1988) 169–215. | MR 968817 | Zbl 0638.60035
[4] . On the functional central limit theorem and the law of the iterated logarithm for Markov processes. Z. Wahrsch. Verw. Gebiete 60 (2) (1982) 185–201. | MR 663900 | Zbl 0468.60034
[5] . Flows of stochastic dynamical systems: Ergodic theory. Stochastics 14 (4) (1985) 273–317. | MR 805125 | Zbl 0536.58019
[6] . Stochastic calculus and degenerate boundary value problems. Ann. Inst. Fourier (Grenoble) 42 (3) (1992) 541–624. | Numdam | MR 1182641 | Zbl 0780.35023
[7] and . Coupling methods for multidimensional diffusion processes. Ann. Probab. 17 (1) (1989) 151–177. | MR 972776 | Zbl 0686.60083
[8] and . Ergodicity for Infinite-Dimensional Systems. London Mathematical Society Lecture Note Series 229. Cambridge Univ. Press, Cambridge, 1996. | MR 1417491 | Zbl 0849.60052
[9] , and . Sample path properties of the stochastic flows. Ann. Probab. 32 (1A) (2004) 1–27. | MR 2040774 | Zbl 1056.60031
[10] . Brownian motions on the homeomorphisms of the plane. Ann. Probab. 9 (2) (1981) 232–254. | MR 606986 | Zbl 0457.60013
[11] . Stochastic Stability of Differential Equations. Monographs and Textbooks on Mechanics of Solids and Fluids: Mechanics and Analysis 7. Sijthoff & Noordhoff, Alphen aan den Rijn, 1980. | MR 600653 | Zbl 0441.60060
[12] . Hypoelliptic second order differential equations. Acta Math. 119 (1967) 147–171. | MR 222474 | Zbl 0156.10701
[13] . The Analysis of Linear Partial Differential Operators 1–4. Grundlehren der Mathematischen Wissenschaften [Fundamental Principles of Mathematical Sciences] 275. Springer, Berlin, 1985. | Zbl 0612.35001
[14] and . A comparison theorem for solutions of stochastic differential equations and its applications. Osaka J. Math. 14 (3) (1977) 619–633. | MR 471082 | Zbl 0376.60065
[15] and . A Second Course in Stochastic Processes. Academic Press, New York, 1981. | MR 611513 | Zbl 0469.60001
[16] and . Numerical Solution of Stochastic Differential Equations. Applications of Mathematics (New York) 23. Springer, Berlin, 1992. | MR 1214374 | Zbl 0752.60043
[17] . Stochastic Flows and Stochastic Differential Equations. Cambridge Studies in Advanced Mathematics 24. Cambridge Univ. Press, Cambridge, 1990. | MR 1070361 | Zbl 0743.60052
[18] and . Recursive computation of the invariant distribution of a diffusion. Bernoulli 8 (3) (2002) 367–405. | MR 1913112 | Zbl 1006.60074
[19] and . Recursive computation of the invariant distribution of a diffusion: The case of a weakly mean reverting drift. Stoch. Dyn. 3 (4) (2003) 435–451. | MR 2030742 | Zbl 1044.60069
[20] . Estimation récursive de la mesure invariante d’un processus de diffusion. Thèse de doctorat, Univ. Marne-la-Vallée, 2005.
[21] , and . Invariant distribution of duplicated diffusions and application to Richardson–Romberg extrapolation, 2014. Extended version available at arXiv:1302.1651. | MR 3414458
[22] . Multi-step Richardson–Romberg extrapolation: Remarks on variance control and complexity. Monte Carlo Methods Appl. 13 (1) (2007) 37–70. | MR 2325670 | Zbl 1119.65004
[23] . Approximation récursive du régime stationnaire d’une Équation Différentielle Stochastique avec sauts. Thèse de doctorat, Univ. Pierre et Marie Curie, 2006.
[24] and . On the Poisson equation and diffusion approximation. I. Ann. Probab. 29 (3) (2001) 1061–1085. | MR 1872736 | Zbl 1029.60053
[25] and . Expansion of the global error for numerical schemes solving stochastic differential equations. Stoch. Anal. Appl. 8 (4) (1990) 483–509. | MR 1091544 | Zbl 0718.60058
[26] . Collapse of attractors for ODEs under small random perturbations. Probab. Theory Related Fields 141 (1–2) (2008) 1–18. | MR 2372963 | Zbl 1143.34034
[27] . Optimal Transport: Old and New. Grundlehren der Mathematischen Wissenschaften [Fundamental Principles of Mathematical Sciences] 338. Springer, Berlin, 2009. | MR 2459454 | Zbl 1156.53003