On Unbiased Estimation for Discretized Models

In this article, we consider computing expectations w.r.t. probability measures which are subject to discretization error. Examples include partially observed diffusion processes or inverse problems, where one may have to discretize time and/or space, in order to practically work with the probability of interest. Given access only to these discretizations, we consider the construction of unbiased Monte Carlo estimators of expectations w.r.t. such target probability distributions. It is shown how to obtain such estimators using a novel adaptation of randomization schemes and Markov simulation methods. Under appropriate assumptions, these estimators possess finite variance and finite expected cost. There are two important consequences of this approach: (i) unbiased inference is achieved at the canonical complexity rate, and (ii) the resulting estimators can be generated independently, thereby allowing strong scaling to arbitrarily many parallel processors. Several algorithms are presented, and applied to some examples of Bayesian inference problems, with both simulated and real observed data.

[1]  L. Rogers,et al.  Coupling of Multidimensional Diffusions by Reflection , 1986 .

[2]  Peter Richtárik,et al.  SGD: General Analysis and Improved Rates , 2019, ICML 2019.

[3]  Andrew M. Stuart,et al.  Inverse problems: A Bayesian perspective , 2010, Acta Numerica.

[4]  A. Eberle,et al.  Coupling and convergence for Hamiltonian Monte Carlo , 2018, The Annals of Applied Probability.

[5]  G. Roberts,et al.  Unbiased Monte Carlo: Posterior estimation for intractable/infinite-dimensional models , 2014, Bernoulli.

[6]  Arnold Neumaier,et al.  Introduction to Numerical Analysis , 2001 .

[7]  Peter W. Glynn,et al.  Unbiased Estimation with Square Root Convergence for SDE Models , 2015, Oper. Res..

[8]  Matti Vihola,et al.  Unbiased Estimators and Multilevel Monte Carlo , 2015, Oper. Res..

[9]  Philippe G. Ciarlet,et al.  The finite element method for elliptic problems , 2002, Classics in applied mathematics.

[10]  A. P. Dawid,et al.  Regression and Classification Using Gaussian Process Priors , 2009 .

[11]  Ajay Jasra,et al.  On Unbiased Score Estimation for Partially Observed Diffusions , 2021, 2105.04912.

[12]  Ajay Jasra,et al.  Unbiased filtering of a class of partially observed diffusions , 2020, Advances in Applied Probability.

[13]  John O'Leary,et al.  Unbiased Markov chain Monte Carlo with couplings , 2017, 1708.03625.

[14]  V. Johnson Studying Convergence of Markov Chain Monte Carlo Algorithms Using Coupled Sample Paths , 1996 .

[15]  Albert Tarantola,et al.  Inverse problem theory - and methods for model parameter estimation , 2004 .

[16]  H. Kushner,et al.  Stochastic Approximation and Recursive Algorithms and Applications , 2003 .

[17]  S. Duane,et al.  Hybrid Monte Carlo , 1987 .

[18]  L. R. Scott,et al.  The Mathematical Theory of Finite Element Methods , 1994 .

[19]  G. Roberts,et al.  MCMC Methods for Functions: ModifyingOld Algorithms to Make Them Faster , 2012, 1202.0709.

[20]  Fredrik Lindsten,et al.  Smoothing With Couplings of Conditional Particle Filters , 2017, Journal of the American Statistical Association.

[21]  J. Heng,et al.  Unbiased Hamiltonian Monte Carlo with couplings , 2017, Biometrika.

[22]  Peter W. Glynn,et al.  Exact estimation for Markov chain equilibrium expectations , 2014, Journal of Applied Probability.

[23]  A. Beskos,et al.  Multilevel sequential Monte Carlo samplers , 2015, 1503.07259.

[24]  Don McLeish,et al.  A general method for debiasing a Monte Carlo estimator , 2010, Monte Carlo Methods Appl..

[25]  Kody J. H. Law,et al.  Unbiased estimation of the gradient of the log-likelihood in inverse problems , 2020, Statistics and Computing.

[26]  Matti Vihola,et al.  Unbiased Inference for Discretely Observed Hidden Markov Model Diffusions , 2018, SIAM/ASA J. Uncertain. Quantification.

[27]  George V. Moustakides,et al.  Extension of Wald's first lemma to Markov processes , 1999, Journal of Applied Probability.

[28]  Yan Zhou,et al.  Multilevel Particle Filters , 2015, SIAM J. Numer. Anal..

[29]  H. Thorisson Coupling, stationarity, and regeneration , 2000 .

[30]  Yan Zhou,et al.  Multilevel Sequential Monte Carlo with Dimension-Independent Likelihood-Informed Proposals , 2017, SIAM/ASA J. Uncertain. Quantification.

[31]  D. Brockmann,et al.  Effective containment explains subexponential growth in recent confirmed COVID-19 cases in China , 2020, Science.

[32]  Ajay Jasra,et al.  Markov chain simulation for multilevel Monte Carlo , 2018, Foundations of Data Science.