Show simple item record

dc.contributor.advisorParker, Kathlyn A.en_US
dc.contributor.authorWan, Huien_US
dc.contributor.otherDepartment of Computer Scienceen_US
dc.date.accessioned2012-05-15T18:07:11Z
dc.date.available2012-05-15T18:07:11Z
dc.date.issued1-Aug-10en_US
dc.date.submittedAug-10en_US
dc.identifierWan_grad.sunysb_0771E_10151.pdfen_US
dc.identifier.urihttp://hdl.handle.net/1951/55662
dc.description.abstractReasoning about uncertainty has been an active area of research for over thirty years. With the advent of networked systems, especially the Web, it became important to be able to combine uncertain information that comes from different sources. One of the hard problems in this area is combining evidence obtained from correlated, possibly conflicting, information sources.This dissertation develops Belief Logic Programming (BLP), a novel form of quantitative logic programming that deals with uncertain and inconsistent information and is able to combine and correlate evidence obtained from non-independent information sources. BLP was inspired by Dempster-Shafer theory of evidence and belief combination functions. Our approach does not depend on a particular method of combining evidence and, in fact, different combination methods can be used simultaneously for different types of uncertain information. Most importantly, unlike previous efforts to integrate uncertainty and logic programming, BLP can correlate structural information contained in rules and provides more accurate estimates for certainty factors.Together with declarative and fixpoint semantics for BLP, the dissertation develops optimized query evaluation algorithms. In addition, the monotonicity and non-monotonicity properties of BLP as well as the relationship to defeasible reasoning, paraconsistent reasoning, and Dempster-Shafer theory of evidence are discussed.After developing the basic framework, the dissertation develops several extensions. One extension allows cyclic dependencies in BLP rules. Another extension captures quantitative correlation among the base facts. Since full correlation information might not always be available, we develop an approach that allows to approximate correlation information using only partial information. Under certain conditions, this approximate method yields the same results as the method that is based on full information. The query evaluation algorithms are then extended to cover the enhancements.en_US
dc.description.sponsorshipStony Brook University Libraries. SBU Graduate School in Department of Computer Science. Lawrence Martin (Dean of Graduate School).en_US
dc.formatElectronic Resourceen_US
dc.language.isoen_USen_US
dc.publisherThe Graduate School, Stony Brook University: Stony Brook, NY.en_US
dc.subject.lcshComputer Scienceen_US
dc.subject.othercorrelation, Dempster-Shafer theory, knowledge representation, logic programming, quantitative reasoning, uncertainty reasoningen_US
dc.titleReasoning about Uncertainty and Correlated Beliefsen_US
dc.typeDissertationen_US
dc.description.advisorAdvisor(s): Michael Kifer. Committee Member(s): C. R. Ramakrishnan; I. V. Ramakrishnan; Yanhong A. Liu; V. S. Subrahmanian.en_US
dc.mimetypeApplication/PDFen_US


Files in this item

Thumbnail

This item appears in the following Collection(s)

Show simple item record