Parallel Markov Chain Monte Carlo

Markov chain Monte Carlo is an inherently serial algorithm. Although likelihood calculations for individual steps can sometimes be parallelized, the serial evolution of the process is widely viewed as incompatible with parallelization, offering no speedup for samplers which require large numbers of iterations to converge to equilibrium. We provide a methodology for parallelizing Markov chain Monte Carlo across large numbers of independent, asynchronous processors. Our approach uses a partitioning and weight estimation scheme to combine independent simulations run on separate processors into rigorous Monte Carlo estimates. The method is originally motivated by sampling multimodal target distributions, where we see an exponential speedup in running time. However we show that the approach is general-purpose and applicable to all Markov chain Monte Carlo simulations, and demonstrate speedups proportional to the number of available processors on slowly mixing chains with unimodal target distributions. The approach is simple and easy to implement, and suggests additional directions for further research.

[1]  Xiao-Li Meng,et al.  Warp Bridge Sampling , 2002 .

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

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

[4]  Manisha Desai,et al.  A new mixture model approach to analyzing allelic-loss data using Bayes factors , 2004, BMC Bioinformatics.

[5]  Jonathan Michael Robert Byrd,et al.  Parallel Markov Chain Monte Carlo , 2010 .

[6]  A. Gelfand.  [Practical Markov Chain Monte Carlo]: Comment , 1992 .

[7]  S. Schmidler,et al.  Adaptive Markov Chain Monte Carlo for Bayesian Variable Selection , 2013 .

[8]  Yanjia Bai.  Divide and Conquer : A Mixture-Based Approach to Regional Adaptation for MCMC , 2009 .

[9]  Sally Rosenthal,et al.  Parallel computing and Monte Carlo algorithms , 1999 .

[10]  D. V. Dyk,et al.  A Bayesian analysis of the multinomial probit model using marginal data augmentation , 2005 .

[11]  Myint Tin.  Comparison of Some Ratio Estimators , 1965 .

[12]  Arnaud Doucet,et al.  On the Utility of Graphics Cards to Perform Massively Parallel Simulation of Advanced Monte Carlo Methods , 2009, Journal of computational and graphical statistics : a joint publication of American Statistical Association, Institute of Mathematical Statistics, Interface Foundation of North America.

[13]  Chao Yang,et al.  Learn From Thy Neighbor: Parallel-Chain and Regional Adaptive MCMC , 2009 .

[14]  Xizhou Feng,et al.  Parallel algorithms for Bayesian phylogenetic inference , 2003, J. Parallel Distributed Comput..

[15]  S. Chib,et al.  Marginal Likelihood From the Metropolis–Hastings Output , 2001 .

[16]  A. Gelfand,et al.  Bayesian Model Choice: Asymptotics and Exact Calculations , 1994 .

[17]  S. Chib.  Marginal Likelihood from the Gibbs Output , 1995 .

[18]  Scott C. Schmidler,et al.  α-Stable Limit Laws for Harmonic Mean Estimators of Marginal Likelihoods , 2012 .

[19]  Agostino Nobile,et al.  A hybrid Markov chain for the Bayesian analysis of the multinomial probit model , 1998, Stat. Comput..

[20]  Yan Bai,et al.  Divide and Conquer: A Mixture-Based Approach to Regional Adaptation for MCMC , 2009 .

[21]  D. Woodard,et al.  Sufficient Conditions for Torpid Mixing of Parallel and Simulated Tempering , 2009 .

[22]  S. Chib,et al.  Bayesian analysis of binary and polychotomous response data , 1993 .

[23]  M. Newton,et al.  Estimating the Integrated Likelihood via Posterior Simulation Using the Harmonic Mean Identity , 2006 .

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