Scaling MCMC Inference and Belief Propagation to Large, Dense Graphical Models

Abstract

SCALING MCMC INFERENCE AND BELIEF PROPAGATION TO LARGE, DENSE GRAPHICAL MODELS MAY 2014 SAMEER SINGH B.E., UNIVERSITY OF DELHI M.Sc., VANDERBILT UNIVERSITY Ph.D., UNIVERSITY OF MASSACHUSETTS AMHERST Directed by: Professor Andrew McCallum With the physical constraints of semiconductor-based electronics becoming increasingly limiting in the past decade, single-core CPUs have given way to multi-core and distributed computing platforms. At the same time, access to large data collections is progressively becoming commonplace due to the lowering cost of storage and bandwidth. Traditional machine learning paradigms that have been designed to operate sequentially on single processor architectures seem destined to become obsolete in this world of multi-core, multi-node systems and massive data sets. Inference for graphical models is one such example for which most existing algorithms are sequential in nature and are difficult to scale using parallel computations. Further, modeling large datasets leads to an escalation in the number of variables, factors, domains, and the density of the models, all of which have a substantial impact on the computational and storage complexity of inference. To achieve scalability, existing techniques impose strict independence assumptions on the model,

56 Figures and Tables

Cite this paper

@inproceedings{Singh2014ScalingMI, title={Scaling MCMC Inference and Belief Propagation to Large, Dense Graphical Models}, author={Sameer Singh and Greg Druck and Kedar Bellare and Pallika Kanani and Laura Dietz and Khashyar Ro and Hanna M. Wallach and Michael Wick and Limin Yao and Jason Narad and Ari Kobren and David Belanger}, year={2014} }