Weighted multilevel Langevin simulation of invariant measures. (English) Zbl 1404.60113

Pagès, Gilles; Panloup, Fabien

Ann. Appl. Probab. 28, No. 6, 3358-3417 (2018).

Summary: We investigate a weighted multilevel Richardson-Romberg extrapolation for the ergodic approximation of invariant distributions of diffusions adapted from the one introduced in [V. Lemaire and the first author, Bernoulli 23, No. 4A, 2643–2692 (2017; Zbl 1383.65003)] for regular Monte Carlo simulation. In a first result, we prove under weak confluence assumptions on the diffusion, that for any integer \(R\geq2\), the procedure allows us to attain a rate \(n^{\frac{R}{2R+1}}\) whereas the original algorithm convergence is at a weak rate \(n^{1/3}\). Furthermore, this is achieved without any explosion of the asymptotic variance. In a second part, under stronger confluence assumptions and with the help of some second-order expansions of the asymptotic error, we go deeper in the study by optimizing the choice of the parameters involved by the method. In particular, for a given \(\varepsilon>0\), we exhibit some semi-explicit parameters for which the number of iterations of the Euler scheme required to attain a mean-squared error lower than \(\varepsilon^{2}\) is about \(\varepsilon^{-2}\log(\varepsilon^{-1})\). { } Finally, we numerically test this multilevel Langevin estimator on several examples including the simple one-dimensional Ornstein-Uhlenbeck process but also a high dimensional diffusion motivated by a statistical problem. These examples confirm the theoretical efficiency of the method.

MSC:

60J60	Diffusion processes
37M25	Computational methods for ergodic theory (approximation of invariant measures, computation of Lyapunov exponents, entropy, etc.)
65C05	Monte Carlo methods

Keywords:

ergodic diffusion; invariant measure; multilevel; ergodicity; Richardson-Romberg; Monte Carlo; PAC-Bayesian

Citations:

Zbl 1383.65003

References:

[1]	Bhattacharya, R. N. (1982). On the functional central limit theorem and the law of the iterated logarithm for Markov processes. Z. Wahrsch. Verw. Gebiete60 185–201. · Zbl 0468.60034
[2]	Billingsley, P. (1978). Ergodic Theory and Information. Robert E. Krieger Publishing Co., Huntington, NY. Reprint of the 1965 original. · Zbl 0141.16702
[3]	Dalalyan, A. S. and Tsybakov, A. B. (2012). Sparse regression learning by aggregation and Langevin Monte-Carlo. J. Comput. System Sci.78 1423–1443. · Zbl 1244.62054
[4]	Duflo, M. (1997). Random Iterative Models. Applications of Mathematics (New York) 34. Springer, Berlin. Translated from the 1990 French original by Stephen S. Wilson and revised by the author. · Zbl 0868.62069
[5]	Frikha, N. (2016). Multi-level stochastic approximation algorithms. Ann. Appl. Probab.26 933–985. · Zbl 1344.93111
[6]	García Trillos, C. A. (2015). A decreasing step method for strongly oscillating stochastic models. Ann. Appl. Probab.25 986–1029. · Zbl 1316.60111
[7]	Gilbarg, D. and Trudinger, N. S. (1983). Elliptic Partial Differential Equations of Second Order, 2nd ed. Grundlehren der Mathematischen Wissenschaften [Fundamental Principles of Mathematical Sciences] 224. Springer, Berlin. · Zbl 0562.35001
[8]	Giles, M. B. (2008). Multilevel Monte Carlo path simulation. Oper. Res.56 607–617. · Zbl 1167.65316
[9]	Giles, M. B. and Szpruch, L. (2014). Antithetic multilevel Monte Carlo estimation for multi-dimensional SDEs without Lévy area simulation. Ann. Appl. Probab.24 1585–1620. · Zbl 1373.65007
[10]	Hall, P. and Heyde, C. C. (1980). Martingale Limit Theory and Its Application. Academic Press, New York. · Zbl 0462.60045
[11]	Krengel, U. (1985). Ergodic Theorems. De Gruyter Studies in Mathematics6. de Gruyter, Berlin. · Zbl 0575.28009
[12]	Lamberton, D. and Pagès, G. (2002). Recursive computation of the invariant distribution of a diffusion. Bernoulli8 367–405. · Zbl 1006.60074
[13]	Lamberton, D. and Pagès, G. (2003). Recursive computation of the invariant distribution of a diffusion: The case of a weakly mean reverting drift. Stoch. Dyn.3 435–451. · Zbl 1044.60069
[14]	Lemaire, V. (2007). Behavior of the Euler scheme with decreasing step in a degenerate situation. ESAIM Probab. Stat.11 236–247. · Zbl 1181.60086
[15]	Lemaire, V. and Pagès, G. (2017). Multilevel Richardson–Romberg extrapolation. Bernoulli23 2643–2692. · Zbl 1383.65003
[16]	Lemaire, V., Pagès, G. and Panloup, F. (2015). Invariant measure of duplicated diffusions and application to Richardson–Romberg extrapolation. Ann. Inst. Henri Poincaré Probab. Stat.51 1562–1596. · Zbl 1329.60281
[17]	Pagès, G. (2007). Multi-step Richardson–Romberg extrapolation: Remarks on variance control and complexity. Monte Carlo Methods Appl.13 37–70.
[18]	Pagès, G. and Panloup, F. (2009). Approximation of the distribution of a stationary Markov process with application to option pricing. Bernoulli15 146–177. · Zbl 1214.60036
[19]	Pagès, G. and Panloup, F. (2014). A mixed-step algorithm for the approximation of the stationary regime of a diffusion. Stochastic Process. Appl.124 522–565. · Zbl 1284.60075
[20]	Pagès, G. and Panloup, F. (2018). Supplement to “Weighted multilevel Langevin simulation of invariant measures.” DOI:10.1214/17-AAP1364SUPP.
[21]	Panloup, F. (2008). Recursive computation of the invariant measure of a stochastic differential equation driven by a Lévy process. Ann. Appl. Probab.18 379–426. · Zbl 1136.60049
[22]	Pardoux, E. and Veretennikov, A. Yu. (2001). On the Poisson equation and diffusion approximation. I. Ann. Probab.29 1061–1085. · Zbl 1029.60053
[23]	Piccioni, M. and Scarlatti, S. (1994). An iterative Monte Carlo scheme for generating Lie group-valued random variables. Adv. in Appl. Probab.26 616–628. · Zbl 0820.68132
[24]	Richardson, L. F. (1911). The approximate arithmetical solution by finite differences of physical problems including differential equations, with an application to the stresses in a masonry dam. Philos. Trans. R. Soc. Lond. Ser. A210 307–357. · JFM 42.0873.02
[25]	Richardson, L. F. (1927). The deferred approach to the limit. Philos. Trans. R. Soc. Lond. Ser. A226 299–349. · JFM 53.0432.01
[26]	Roberts, G. O. and Rosenthal, J. S. (1998). Optimal scaling of discrete approximations to Langevin diffusions. J. R. Stat. Soc. Ser. B. Stat. Methodol.60 255–268. · Zbl 0913.60060
[27]	Roberts, G. O. and Tweedie, R. L. (1996). Exponential convergence of Langevin distributions and their discrete approximations. Bernoulli2 341–363. · Zbl 0870.60027
[28]	Szpruch, L., Vollmer, S., Zygalakis, K. and Giles, M. B. (2016). Multi Level Monte Carlo methods for a class of ergodic stochastic differential equations. ArXiv e-print. Available at arXiv:1605.01384 [math.NA].
[29]	Talay, D. (1990). Second order discretization schemes of stochastic differential systems for the computation of the invariant law. Stoch. Stoch. Rep.29 13–36. · Zbl 0697.60066