Tue, Dec 07, 2010 @ 10:00 AM - 11:00 AM
Ming Hsieh Department of Electrical and Computer Engineering
Conferences, Lectures, & Seminars
Speaker: Dr. Paul Fodor, Ph.D.
Talk Title: Practical Reasoning with Transaction Logic Programming for Knowledge Base Dynamics
Abstract: Abstract: Transaction Logic is an extension of classical predicate calculus for representing declarative and procedural knowledge in logic programming, databases, and artificial intelligence. Since it provides a logical foundation for the phenomenon of state changes, it has been successful in areas as diverse as workflows, planning, reasoning about actions, Web services, security policies, active databases and more. Although a number of implementations of Transaction Logic exist, none is logically complete due to the time and space complexity of such implementations. In the first part of this talk, I describe an approach for performing actions in the logic, which has better complexity and termination properties via a logically complete tabling evaluation strategy. Then I describe a series of optimizations, which make this algorithm practical and analyze their performance on a set of benchmarks. Our performance evaluation study shows that the tabling algorithm can scale well both in time and space. Appart from applications in classical logic programming use cases we used the new system for applications in the complex event processing domain. In the second part of the talk, I describe our extension of Transaction Logic in the direction of defeasible reasoning, which has a number of interesting applications, including specification of defaults in action theories and heuristics for directed search in planning. In this setting we showed that heuristics expressed as defeasible actions can significantly reduce the search space and thus the execution time and space requirements.
Bio: Paul Fodor obtained a Bachelor and a Master of Engineering in Computer Science from the Technical University of Cluj-Napoca, Romania, a M.Sc. from the Computer Science Department, State University of New York at Stony Brook in 2006 and is expected to graduate with Ph.D. in 2010. His PhD has resulted in multiple peer-reviewed publications and presentations in conferences such as WWW, ICLP, PPDP, Web Rules and Reasoning, DEBS, AAAI, PADL, ACL, and LTC. He has supplemented his academic research with industrial experience by working as a developer at DaimlerChrysler AG and IBM and as a research intern for three summers at IBM. In IBM TJ Watson Research he worked on the Watson NLP system scheduled to play on the Jeopardy! TV quiz show. As a Ph.D. student, Paul worked on declarative rule languages and logic used as a specification language and implementation framework for active knowledge bases. Through his work, Paul also contributed to several large software projects: OpenRuleBench (a suite of benchmarks for analyzing the performance and scalability of rule systems for the semantic Web), Event Transaction Logic Inference System - ETALIS (a declarative complex event processing language), Semantic Inferencing on Large Knowledge - SILK (a multi-institution project sponsored by Vulcan Inc. aiming to provide knowledge representation infrastructure for very large knowledge bases), and Unstructured Information Management Architecture UIMA (an industrial-strength, scalable and extensible platform for creating, integrating and deploying unstructured information management systems).
Host: Professor Viktor K. Prasanna
Audiences: Everyone Is Invited
Contact: Janice Thompson