Noisy Monte Carlo: convergence of Markov chains with approximate transition kernels

DOI10.1007/S11222-014-9521-XzbMATH Open1342.60122arXiv1403.5496OpenAlexW1964607942MaRDI QIDQ2631344FDOQ2631344

Authors: Pierre Alquier, Nial Friel, Richard G. Everitt, Aidan Boland 

Publication date: 29 July 2016

Published in: Statistics and Computing (Search for Journal in Brave)

Abstract: Monte Carlo algorithms often aim to draw from a distribution $pi$ by simulating a Markov chain with transition kernel $P$ such that $pi$ is invariant under $P$. However, there are many situations for which it is impractical or impossible to draw from the transition kernel $P$. For instance, this is the case with massive datasets, where is it prohibitively expensive to calculate the likelihood and is also the case for intractable likelihood models arising from, for example, Gibbs random fields, such as those found in spatial statistics and network analysis. A natural approach in these cases is to replace $P$ by an approximation $hatP$. Using theory from the stability of Markov chains we explore a variety of situations where it is possible to quantify how 'close' the chain given by the transition kernel $hatP$ is to the chain given by $P$. We apply these results to several examples from spatial statistics and network analysis.

Full work available at URL: https://arxiv.org/abs/1403.5496

Recommendations

• Markov Chain Monte Carlo Algorithms: Theory and Practice
• On nonlinear Markov chain Monte Carlo
• Approximations of geometrically ergodic reversible Markov chains
• Using a Markov Chain to Construct a Tractable Approximation of an Intractable Probability Distribution
• Stability of noisy Metropolis-Hastings

zbMATH Keywords

Markov chain Monte Carlointractable likelihoodspseudo-marginal Monte Carlo

Mathematics Subject Classification ID

Convergence of probability measures (60B10) Directional data; spatial statistics (62H11) Monte Carlo methods (65C05) Markov chains (discrete-time Markov processes on discrete state spaces) (60J10) Interacting random processes; statistical mechanics type models; percolation theory (60K35)

Cites Work

• Title not available (Why is that?)
• Statistics for high-dimensional data. Methods, theory and applications.
• Exponential convergence of Langevin distributions and their discrete approximations
• Title not available (Why is that?)
• Title not available (Why is that?)
• The pseudo-marginal approach for efficient Monte Carlo computations
• Stochastic Relaxation, Gibbs Distributions, and the Bayesian Restoration of Images
• A Stochastic Approximation Method
• Approximate Bayesian computational methods
• Langevin diffusions and Metropolis-Hastings algorithms
• Geometric convergence and central limit theorems for multidimensional Hastings and Metropolis algorithms
• Markov chains and stochastic stability
• Riemann manifold Langevin and Hamiltonian Monte Carlo methods. With discussion and authors' reply
• An efficient Markov chain Monte Carlo method for distributions with intractable normalising constants
• Sensitivity and convergence of uniformly ergodic Markov chains
• A theory of the learnable
• Exact sampling with coupled Markov chains and applications to statistical mechanics
• Convergence properties of pseudo-marginal Markov chain Monte Carlo algorithms
• Regular Perturbation of V-Geometrically Ergodic Markov Chains
• Sparse regression learning by aggregation and Langevin Monte-Carlo
• Recursive computing and simulation-free inference for general factorizable models
• Title not available (Why is that?)
• Efficient recursions for general factorisable models

Cited In (48)

• Parallel sequential Monte Carlo for stochastic gradient-free nonconvex optimization
• Perturbation bounds for Monte Carlo within metropolis via restricted approximations
• Subsampling MCMC -- an introduction for the survey statistician
• Semiparametric Bayesian analysis for longitudinal mixed effects models with non-normal AR(1) errors
• Markov Kernels Local Aggregation for Noise Vanishing Distribution Sampling
• Ensemble Kalman methods for high-dimensional hierarchical dynamic space-time models
• Informed sub-sampling MCMC: approximate Bayesian inference for large datasets
• A Bayesian multilevel model for populations of networks using exponential-family random graphs
• Perturbation and Inverse Problems of Stochastic Matrices
• A rare event approach to high-dimensional approximate Bayesian computation
• Sequential Monte Carlo with transformations
• Model comparison for Gibbs random fields using noisy reversible jump Markov chain Monte Carlo
• Accelerating pseudo-marginal MCMC using Gaussian processes
• Bayesian model selection for high-dimensional Ising models, with applications to educational data
• Sequential tests for large-scale learning
• Noisy Hamiltonian Monte Carlo for Doubly Intractable Distributions
• Using a Markov Chain to Construct a Tractable Approximation of an Intractable Probability Distribution
• Computationally efficient inference for latent position network models
• Uncertainty quantification for Markov processes via variational principles and functional inequalities
• A multilayer exponential random graph modelling approach for weighted networks
• Robustness of iterated function systems of Lipschitz maps
• Bayesian computation: a summary of the current state, and samples backwards and forwards
• Stability of doubly-intractable distributions
• Perturbation theory for Markov chains via Wasserstein distance
• Bayesian inference in the presence of intractable normalizing functions
• The Monte Carlo computation error of transition probabilities
• An efficient adaptive MCMC algorithm for pseudo-Bayesian quantum tomography
• A Function Emulation Approach for Doubly Intractable Distributions
• User-friendly guarantees for the Langevin Monte Carlo with inaccurate gradient
• Monte Carlo Markov chains constrained on graphs for a target with disconnected support
• Stability of noisy Metropolis-Hastings
• Bayesian inference, model selection and likelihood estimation using fast rejection sampling: the Conway-Maxwell-Poisson distribution
• Approximations of geometrically ergodic reversible Markov chains
• Estimating promotion effects in email marketing using a large-scale cross-classified Bayesian joint model for nested imbalanced data
• Bayesian model comparison with un-normalised likelihoods
• Title not available (Why is that?)
• Multivariate Conway-Maxwell-Poisson Distribution: Sarmanov Method and Doubly Intractable Bayesian Inference
• On coupling particle filter trajectories
• A Bayesian approach to disease clustering using restricted Chinese restaurant processes
• Likelihood-free approximate Gibbs sampling
• An algorithm for distributed Bayesian inference
• Scalable Bayes via barycenter in Wasserstein space
• Efficient MCMC for Gibbs random fields using pre-computation
• MEXIT: maximal un-coupling times for stochastic processes
• Exploiting multi-core architectures for reduced-variance estimation with intractable likelihoods
• Distributed computation for marginal likelihood based model choice
• On Russian roulette estimates for Bayesian inference with doubly-intractable likelihoods
• Distributed Bayesian Inference in Linear Mixed-Effects Models

This page was built for publication: Noisy Monte Carlo: convergence of Markov chains with approximate transition kernels

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q2631344)