A Monte Carlo Metropolis-Hastings Algorithm for Sampling from Distributions with Intractable Normalizing Constants

Simulating from distributions with intractable normalizing constants has been a long-standing problem in machine learning. In this letter, we propose a new algorithm, the Monte Carlo Metropolis-Hastings (MCMH) algorithm, for tackling this problem. The MCMH algorithm is a Monte Carlo version of the Metropolis-Hastings algorithm. It replaces the unknown normalizing constant ratio by a Monte Carlo estimate in simulations, while still converges, as shown in the letter, to the desired target distribution under mild conditions. The MCMH algorithm is illustrated with spatial autologistic models and exponential random graph models. Unlike other auxiliary variable Markov chain Monte Carlo (MCMC) algorithms, such as the Møller and exchange algorithms, the MCMH algorithm avoids the requirement for perfect sampling, and thus can be applied to many statistical models for which perfect sampling is not available or very expensive. The MCMH algorithm can also be applied to Bayesian inference for random effect models and missing data problems that involve simulations from a distribution with intractable integrals.

[1]  Tom A. B. Snijders,et al.  A comparison of various approaches to the exponential random graph model: A reanalysis of 102 student networks in school classes , 2007, Soc. Networks.

[2]  Tatiyana V. Apanasovich,et al.  On estimation in binary autologistic spatial models , 2006 .

[3]  S. Wasserman,et al.  Models and Methods in Social Network Analysis , 2005 .

[4]  M. Kenward,et al.  An Introduction to the Bootstrap , 2007 .

[5]  Ming-Hui Chen,et al.  On Monte Carlo methods for estimating ratios of normalizing constants , 1997 .

[6]  Jun S. Liu,et al.  Monte Carlo Maximum Likelihood for Exponential Random Graph Models: From Snowballs to Umbrella Densities , 2009 .

[7]  S. Wasserman,et al.  Models and Methods in Social Network Analysis: An Introduction to Random Graphs, Dependence Graphs, and p * , 2005 .

[8]  C. Geyer,et al.  Constrained Monte Carlo Maximum Likelihood for Dependent Data , 1992 .

[9]  Zoubin Ghahramani,et al.  MCMC for Doubly-intractable Distributions , 2006, UAI.

[10]  D. J. Strauss,et al.  Pseudolikelihood Estimation for Social Networks , 1990 .

[11]  Garry Robins,et al.  An introduction to exponential random graph (p*) models for social networks , 2007, Soc. Networks.

[12]  D. Rubin,et al.  Inference from Iterative Simulation Using Multiple Sequences , 1992 .

[13]  P. Pattison,et al.  New Specifications for Exponential Random Graph Models , 2006 .

[14]  Faming Liang,et al.  An Overview of Stochastic Approximation Monte Carlo , 2014 .

[15]  Mark Von Tress,et al.  Generalized, Linear, and Mixed Models , 2003, Technometrics.

[16]  David Bruce Wilson,et al.  Exact sampling with coupled Markov chains and applications to statistical mechanics , 1996, Random Struct. Algorithms.

[17]  P. Bickel,et al.  Mathematical Statistics: Basic Ideas and Selected Topics , 1977 .

[18]  Håvard Rue,et al.  A Tutorial on Image Analysis , 2003 .

[19]  M. Wu,et al.  Modeling the Distribution of Plant Species Using the Autologistic Regression Model Modeling the Distribution of Plant Species Using the Autologistic Regression Model , 1997 .

[20]  Hulin Wu,et al.  Modelling the distribution of plant species using the autologistic regression model , 1997, Environmental and Ecological Statistics.

[21]  L. Tierney.  Markov Chains for Exploring Posterior Distributions , 1994 .

[22]  Tom A. B. Snijders,et al.  Markov Chain Monte Carlo Estimation of Exponential Random Graph Models , 2002, J. Soc. Struct..

[23]  A. Childs,et al.  Exact sampling from nonattractive distributions using summary states. , 2000, Physical review. E, Statistical, nonlinear, and soft matter physics.

[24]  J. Møller,et al.  An efficient Markov chain Monte Carlo method for distributions with intractable normalising constants , 2006 .

[25]  Neal Madras.  Lectures on Monte Carlo Methods , 2002 .

[26]  D. Hunter,et al.  Inference in Curved Exponential Family Models for Networks , 2006 .

[27]  David R. Hunter,et al.  Curved exponential family models for social networks , 2007, Soc. Networks.

[28]  D. Hunter,et al.  Goodness of Fit of Social Network Models , 2008 .

[29]  C. Geyer.  Markov Chain Monte Carlo Maximum Likelihood , 1991 .

[30]  Xiao-Li Meng,et al.  SIMULATING RATIOS OF NORMALIZING CONSTANTS VIA A SIMPLE IDENTITY: A THEORETICAL EXPLORATION , 1996 .

[31]  Martina Morris,et al.  Specification of Exponential-Family Random Graph Models: Terms and Computational Aspects. , 2008, Journal of statistical software.

[32]  J.-P. Dubois,et al.  Assessing the risk of soil contamination in the Swiss Jura using indicator geostatistics , 1997, Environmental and Ecological Statistics.

[33]  Leonhard Held,et al.  Gaussian Markov Random Fields: Theory and Applications , 2005 .

[34]  Faming Liang,et al.  A double Metropolis–Hastings sampler for spatial models with intractable normalizing constants , 2010 .

[35]  Peter Shawn Bearman,et al.  New methods for new research on adolescent sexual behavior , 1998 .

[36]  Xiao-Li Meng,et al.  Simulating Normalizing Constants: From Importance Sampling to Bridge Sampling to Path Sampling , 1998 .

[37]  Neal Madras,et al.  Monte Carlo methods , 2000 .

[38]  Mark S. Handcock,et al.  A framework for the comparison of maximum pseudo-likelihood and maximum likelihood estimation of exponential family random graph models , 2009, Soc. Networks.

[39]  G. C. Wei,et al.  A Monte Carlo Implementation of the EM Algorithm and the Poor Man's Data Augmentation Algorithms , 1990 .

[40]  G. Torrie,et al.  Nonphysical sampling distributions in Monte Carlo free-energy estimation: Umbrella sampling , 1977 .

[41]  D. Rubin,et al.  Maximum likelihood from incomplete data via the EM - algorithm plus discussions on the paper , 1977 .

[42]  Ove Frank,et al.  http://www.jstor.org/about/terms.html. JSTOR's Terms and Conditions of Use provides, in part, that unless you have obtained , 2007 .

[43]  F. Liang.  Continuous Contour Monte Carlo for Marginal Density Estimation With an Application to a Spatial Statistical Model , 2007 .

[44]  R. Jessor,et al.  New perspectives on adolescent risk behavior: Author index , 1998 .

[45]  Qian Chen,et al.  Approach for LIDAR signals with multiple returns. , 2014, Applied optics.

[46]  Jenine K. Harris.  An Introduction to Exponential Random Graph Modeling , 2013 .

[47]  J. Besag.  Spatial Interaction and the Statistical Analysis of Lattice Systems , 1974 .

[48]  P. Green,et al.  Hidden Markov Models and Disease Mapping , 2002 .

[49]  C. Andrieu,et al.  The pseudo-marginal approach for efficient Monte Carlo computations , 2009, 0903.5480.

[50]  M. Beaumont.  Estimation of population growth or decline in genetically monitored populations. , 2003, Genetics.

[51]  H. Preisler,et al.  Modelling Spatial Patterns of Trees Attacked by Bark-beetles , 1993 .

[52]  Trish,et al.  Protecting adolescents from harm. Findings from the National Longitudinal Study on Adolescent Health. , 1997, JAMA.