On Russian Roulette Estimates for Bayesian Inference with Doubly-Intractable Likelihoods

Abstract

A large number of statistical models are "doubly-intractable": the likelihood normalising term, which is a function of the model parameters, is intractable, as well as the marginal likelihood (model evidence). This means that standard inference techniques to sample from the posterior, such as Markov chain Monte Carlo (MCMC), cannot be used. Examples include, but are not confined to, massive Gaussian Markov random fields, autologistic models and Exponential random graph models. A number of approximate schemes based on MCMC techniques, Approximate Bayesian computation (ABC) or analytic approximations to the posterior have been suggested, and these are reviewed here. Exact MCMC schemes, which can be applied to a subset of doubly-intractable distributions, have also been developed and are described in this paper. As yet, no general method exists which can be applied to all classes of models with doubly-intractable posteriors. In addition, taking inspiration from the Physics literature, we study an alternative method based on representing the intractable likelihood as an infinite series. Unbiased estimates of the likelihood can then be obtained by finite time stochastic truncation of the series via Russian Roulette sampling, although the estimates are not necessarily positive. Results from the Quantum Chromodynamics literature are exploited to allow the use of possibly negative estimates in a pseudo-marginal MCMC scheme such that expectations with respect to the posterior distribution are preserved. The methodology is reviewed on well-known examples such as the parameters in Ising models, the posterior for Fisher-Bingham distributions on the $d$-Sphere and a large-scale Gaussian Markov Random Field model describing the Ozone Column data. This leads to a critical assessment of the strengths and weaknesses of the methodology with pointers to ongoing research.

Published at http://dx.doi.org/10.1214/15-STS523 in the Statistical Science (http://www.imstat.org/sts/) by the Institute of Mathematical Statistics (http://www.imstat.org)

References in corpus (9)

• Computational complexity and fundamental limitations to fermionic quantum Monte Carlo simulations
• The pseudo-marginal approach for efficient Monte Carlo computations
• MCMC for doubly-intractable distributions
• On the efficiency of pseudo-marginal random walk Metropolis algorithms
• Nonstationary covariance models for global data
• Spatial models generated by nested stochastic partial differential equations, with an application to global ozone mapping
• Pre-processing for approximate Bayesian computation in image analysis
• Establishing some order amongst exact approximations of MCMCs
• Noisy Monte Carlo: Convergence of Markov chains with approximate transition kernels

Cited by in corpus (40)

• Speeding Up MCMC by Efficient Data Subsampling
• Exponential-Family Models of Random Graphs: Inference in Finite-, Super-, and Infinite Population Scenarios
• Piecewise Deterministic Markov Processes for Continuous-Time Monte Carlo
• Speeding Up MCMC by Delayed Acceptance and Data Subsampling
• An unbiased ray-marching transmittance estimator
• Importance sampling type estimators based on approximate marginal MCMC
• Stochastic Modelling of Urban Structure
• Relaxing Bijectivity Constraints with Continuously Indexed Normalising Flows
• Bayesian computation: a perspective on the current state, and sampling backwards and forwards
• Hamiltonian Monte Carlo with Energy Conserving Subsampling
• Enabling scalable stochastic gradient-based inference for Gaussian processes by employing the Unbiased LInear System SolvEr (ULISSE)
• Computing Bayes: From Then 'Til Now'
• Noisy Hamiltonian Monte Carlo for doubly-intractable distributions
• Distributed Computation for Marginal Likelihood based Model Choice
• On the Pitfalls of Nested Monte Carlo
• Multifidelity Approximate Bayesian Computation with Sequential Monte Carlo Parameter Sampling
• The block-Poisson estimator for optimally tuned exact subsampling MCMC
• Model comparison for Gibbs random fields using noisy reversible jump Markov chain Monte Carlo
• A numerically stable algorithm for integrating Bayesian models using Markov melding
• Estimating Reciprocal Partition Functions to Enable Design Space Sampling
• Stability of Noisy Metropolis-Hastings
• A determinant-free method to simulate the parameters of large Gaussian fields
• Bias-Free Scalable Gaussian Processes via Randomized Truncations
• Subsampling MCMC - An introduction for the survey statistician
• Exact Bayesian Inference for the Bingham Distribution
• Generalized Bayes Approach to Inverse Problems with Model Misspecification
• Unbiased Bayesian Inference for Population Markov Jump Processes via Random Truncations
• Bayesian Analysis of High-dimensional Discrete Graphical Models
• Bring the noise: exact inference from noisy simulations in collider physics
• Entropy-based adaptive Hamiltonian Monte Carlo
• Multilevel Monte Carlo Variational Inference
• Tightening Bounds for Variational Inference by Revisiting Perturbation Theory
• A Conway-Maxwell-Poisson GARMA Model for Count Data
• Estimation and uncertainty quantification for the output from quantum simulators
• Some comments about "Penalising model component complexity" by Simpson et al. (2017)
• Exploiting Multi-Core Architectures for Reduced-Variance Estimation with Intractable Likelihoods
• On MCMC for variationally sparse Gaussian processes: A pseudo-marginal approach
• Measuring Sample Quality in Algorithms for Intractable Normalizing Function Problems
• Optimal Unbiased Estimation for Expected Cumulative Cost
• Reward Learning with Intractable Normalizing Functions