Electronic Theses and Dissertations

Identifier

6220

Date

2018

Document Type

Thesis

Degree Name

Master of Science

Major

Computer Science

Committee Chair

Deepak Venugopal

Committee Member

Lan Wang

Committee Member

Scott Fleming

Abstract

Rao-Blackwellisation is a technique that provably improves the performance of Gibbs sampling by summing-out variables from the PGM. However, collapsing variables is computationally expensive, since it changes the PGM structure introducing factors whose size is dependent upon the Markov blanket of the variable. Therefore, collapsing out several variables jointly is typically intractable in arbitrary PGM structures. This thesis proposes an adaptive approach for Rao-Blackwellisation, where additional parallel Markov chains are defined over different collapsed PGM structures. The collapsed variables are chosen based on their convergence diagnostics. Adding chains requires re-burn-in the chain, thus wasting samples. To address this, new chains are initialized from a mean field approximation for the distribution, that improves over time, thus reducing the burn-in period. The experiments on several UAI benchmarks shows that this approach is more accurate than state-of-the-art inference systems such as Merlin which have previously won the UAI inference challenge.

Comments

Data is provided by the student.

Library Comment

Dissertation or thesis originally submitted to the local University of Memphis Electronic Theses & dissertation (ETD) Repository.

Download

Share

COinS