From Fields to Trees

We present new MCMC algorithms for computing the posterior distributions and expectations of the unknown variables in undirected graphical models with regular structure. For demonstration purposes, we focus on Markov Random Fields (MRFs). By partitioning the MRFs into non-overlapping trees, it is possible to compute the posterior distribution of a particular tree exactly by conditioning on the remaining tree. These exact solutions allow us to construct efficient blocked and Rao-Blackwellised MCMC algorithms. We show empirically that tree sampling is considerably more efficient than other partitioned sampling schemes and the naive Gibbs sampler, even in cases where loopy belief propagation fails to converge. We prove that tree sampling exhibits lower variance than the naive Gibbs sampler and other naive partitioning schemes using the theoretical measure of maximal correlation. We also construct new information theory tools for comparing different MCMC schemes and show that, under these, tree sampling is more efficient.

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

[2]  Donald Geman,et al.  Stochastic Relaxation, Gibbs Distributions, and the Bayesian Restoration of Images , 1984, IEEE Transactions on Pattern Analysis and Machine Intelligence.

[3]  J. Besag.  On the Statistical Analysis of Dirty Pictures , 1986 .

[4]  Wang,et al.  Nonuniversal critical dynamics in Monte Carlo simulations. , 1987, Physical review letters.

[5]  Judea Pearl,et al.  Evidential Reasoning Using Stochastic Simulation of Causal Models , 1987, Artif. Intell..

[6]  Adrian F. M. Smith,et al.  Sampling-Based Approaches to Calculating Marginal Densities , 1990 .

[7]  Jun S. Liu,et al.  Covariance structure of the Gibbs sampler with applications to the comparisons of estimators and augmentation schemes , 1994 .

[8]  R. Kohn,et al.  On Gibbs sampling for state space models , 1994 .

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

[10]  W. Wefelmeyer,et al.  Outperforming the Gibbs sampler empirical estimator for nearest-neighbor random fields , 1996 .

[11]  G. Casella,et al.  Rao-Blackwellisation of sampling schemes , 1996 .

[12]  W. Freeman,et al.  Generalized Belief Propagation , 2000, NIPS.

[13]  Nando de Freitas,et al.  Rao-Blackwellised Particle Filtering for Dynamic Bayesian Networks , 2000, UAI.

[14]  Jun S. Liu,et al.  Monte Carlo strategies in scientific computing , 2001 .

[15]  Stan Z. Li,et al.  Markov Random Field Modeling in Image Analysis , 2001, Computer Science Workbench.

[16]  Darren J. Wilkinson,et al.  Conditional simulation from highly structured Gaussian systems, with application to blocking-MCMC for the Bayesian analysis of very large linear models , 2002, Stat. Comput..

[17]  Martial Hebert,et al.  Discriminative Fields for Modeling Spatial Dependencies in Natural Images , 2003, NIPS.

[18]  Rina Dechter,et al.  An Empirical Study of w-Cutset Sampling for Bayesian Networks , 2003, UAI.

[19]  Martin J. Wainwright,et al.  Tree-reweighted belief propagation algorithms and approximate ML estimation by pseudo-moment matching , 2003, AISTATS.

[20]  Mark A. Paskin.  Sample Propagation , 2003, NIPS.

[21]  Andrew McCallum,et al.  Dynamic Conditional Random Fields for Jointly Labeling Multiple Sequences , 2003 .

[22]  Christian P. Robert,et al.  Monte Carlo Statistical Methods , 2005, Springer Texts in Statistics.

[23]  Christian P. Robert,et al.  Monte Carlo Statistical Methods (Springer Texts in Statistics) , 2005 .