Main content

CPT Approximation with NIN-AND Tree Causal Models

Show simple item record

dc.contributor.advisor Xiang, Yang Liu, Qing 2014-01-30T18:37:31Z 2014-01-30T18:37:31Z 2014-01 2014-01-24 2014-01-30
dc.description.abstract A Bayesian network (BN) is a probabilistic graphical model widely used in artificial intelligence (AI) to support uncertain knowledge representation and uncertain reasoning. In BNs, conditional probability tables (CPTs) are used to store quantitative knowledge. The size of a discrete variable’s CPT increases exponentially on the number of related causes. Causal models are proposed to specify a CPT with fewer parameters. Non-Impeding Noisy-AND tree (NIN-AND tree) causal modeling is an expressive causal model which can encode two types of causal interactions. Replacing a CPT with a NIN-AND tree causal model can save storage space, and speed up inference. Being motivated by the advantages of using NIN-AND tree causal modeling, we develop the techniques to approximate a given CPT with NIN-AND tree causal modeling in this research. In particular, we develop a suite of algorithms to approximate an arbitrary CPT of binary variables with a NIN-AND tree model. Based on the experimental results, the methods proposed in this research can result in reasonably good approximation accuracy. en_US
dc.description.sponsorship Xiang Yang en_US
dc.language.iso en en_US
dc.subject CPT approximation en_US
dc.title CPT Approximation with NIN-AND Tree Causal Models en_US
dc.type Thesis en_US Computer Science en_US Master of Science en_US School of Computer Science en_US
dc.rights.license All items in the Atrium are protected by copyright with all rights reserved unless otherwise indicated.

Files in this item

Files Size Format View Description
Liu_Qing_201401_Msc.pdf.pdf 1.271Mb PDF View/Open Qing's Thesis

This item appears in the following Collection(s)

Show simple item record