Hosted By
Department of Computer Science
University College, The University of New South Wales
Australian Defence Force Academy, Canberra, ACT 2600, Australia
Last modified on Wed Jul 5 09:38:15 1995
Complete information is difficult to come by and generally not available even in simple database applications. Consequently, an intelligent system must be capable of making plausible conjectures which may be retracted when found to be incorrect according to new information that becomes available. Nonmonotonic Reasoning provides mechanisms for an intelligent reasoning system to make such conjectures when its knowledge is incomplete. This is accomplished in several ways, for example by using rules of thumb (default logic) or introspection (autoepistemic logic).
The aim of this tutorial is to present the basic ideas and approaches to nonmonotonic reasoning. Emphasis will be placed on providing operatioal characterizations of the logics to be discussed, in order to demonstrate that they can applied to concrete examples in a straightforward way. Furthermore, we shall show how one can represent problems in domains such as law or diagnosis.
The course will be based on material of a textbook currently under preparation, which will be published by The MIT Press in the early 1996.
Since intelligent systems are usually faced with incomplete information, nonmonotonic reasoning is a central issue for many artificial intelligence applications in domains such as legal reasoning, diagnosis, or natural language understanding. The participants will obtain an appreciation of some methods in the field, and will explore tools and techniques which can be very useful in practice.
1. Nonmonotonic Reasoning: What is it all about? 2. Default Logic 2.1 Default rules and theories 2.2 Extensions 2.3 Computing extensions 2.4 A prototype implementation in Prolog 2.5 Normal default theories 2.6 Priorities among defaults 3. Autoepistemic Logic 3.1 The language of autoepistemic logic 3.2 Expansions 3.3 Computing expansions 3.4 A prototype implementation in Prolog 3.5 Connections with Default Logic 4. Nonmonotonic Inference Relations
Grigoris Antoniou studied computer science in Karlsruhe(Germany) and earned his PhD in Osnabruck(Germany). Currently he is a lecturer in Information Systems at the University of Newcastle. His research interests are in Nonmonotonic Reasoning, Knowledge Representation, Logic Programming, and the Logical Foundations of Computer Science and AI. He is co-author of the book "Logic: A Foundation for Computer Science", Addison-Wesley 1991. He has published numerous articles in journals (such as Artificial Intelligence Review, Artificial Intelligence Tools, Expert Systems with Applications, Annals of Mathematics and Artificial Intelligence) and conferences (such as IJCAI and AAAI). A book of his on Nonmonotonic Reasoning will be published by the MIT Press in early 1996. He has extensive teaching experience in AI, and gave tutorials on Nonmonotonic Reasoning at the German Spring School on AI (KIFS-93), and at AI'94 (together with Mary-Anne Williams).