8th Intl. Workshop on Non-Monotonic Reasoning NMR'2000
April 9-11, Breckenridge, Colorado
(Collocated with KR'2000, April 12-15)
 

---

• Program
• List of technical contributions
• Schedule of talks
• Proceedings on line
• Didier Dubois - Invited Talk Slides (NEW)
• Practical information
• Instructions for the authors
• Accommodation
• Registration Information
• Travel info
• General information
• Special sessions
• Abductive reasoning
• Belief change: theory and practice
• Representing actions and planning
• System descriptions and demonstration
• Uncertainty frameworks in NMR
• Submission details
• Important dates
• Organizers and Program Committee

NMR2000 have reserved a block of rooms at The Village at Breckenridge at reduced conference rates from Saturday night, 8 April 2000 through Monday night, 10 April 2000. Rooms are also available at the conference rate from 11 April 2000 through 20 April 2000. To qualify for these rates, reservations must be made by contacting the hotel directly and identifying yourself as a `nonmonotonic workshop' attendee. (For the days after NMR, you may have to identify yourself as KR/AIPS attendee.)

Room availability is limited so attendees are encouraged  to reserve their rooms by 22 FEBRUARY 2000. (Note: Cut-off date for the special price is March 9th.) All reservation requests must be accompanied by, or followed within 10 days of booking by, a first night room deposit including tax. Personal/company checks or credit cards are acceptable forms for deposit funds. The hotel will not hold any reservations unless guaranteed by one of the above methods.

PLEASE NOTE: Early departures are not permitted without penalty of full payment.

Contact information for the conference hotel is:

 The Village at Breckenridge
 P. O. Box 8329
 535 South Park Ave.
 Breckenridge, CO 80424 U. S. A.
 Phone (from North America): 1 (877) 428-7829
 Phone (from other locations): +1 970 453-2000

Conference rates at The Village at Breckenridge (plus taxes of about 11%):
 
 
 

 Check-in time: 4:00 pm
 Check-out time: Prior to 10:00 am
 Distance to conference center: adjacent (BML 2-3 blocks)
 Cut-off date for reservations: 5:00 PM MDT, March 9, 2000

NMR2000 is one of the workshops associated with KR2000.  To register to NMR, please use the KR/AIPS registration form.  NMR2000 workshop is identified there as workshop KW1. The registration information and forms can be obtained from the KR2000 home page at http://www.kr.org/kr/kr00/

On-site registration will be available on Sunday, April 9, from 7:30 am. However, EARLY REGISTRATION IS STRONGLY RECOMMENDED.

Back to the top

Probably the easiest way to travel to Breckenridge is to fly into Denver International Airport and take ground transportation from there. Ground transportation via van shuttle can be arranged from Report Express at +1 970 468-7600 or www.resort-express.com. The current fare is US$45 each way. All major car rental companies rent cars at Denver International Airport.

Information about Breckenridge

The town of Breckenridge is located next to the Breckenridge ski area in the heart of Summit County, Colorado, about 90 miles from Denver. The town is at elevation about 9500 feet (2900 meters). Summit County is known for its skiing but has many other activities and attractions and great natural beauty. There are four major ski areas in the vicinity of  Breckenridge: Arapahoe Basin, Copper Mountain, Breckenridge itself, and Keystone.

The weather in April there is cold, with highs averaging 48F (9C) and lows averaging 17F (-8C). Expect lots of snow, as April is the snowiest month in Summit County.
Back to the top

In mentioning the Village at Breckenridge and all other service providers for the NMR 2000 workshop acts only in the capacity of agent for the service providers which are the provider of hotel rooms, transportation, excursions, and banquets. Because NMR2000 has no control over the personnel, equipment or operations of service providers of accommodations or other services included as part of the conference program, NMR2000 assumes no responsibilityfor and will not be liable for any personal delay, inconveniences or other damage suffered by conference participants which may arise by reason of (1) any wrongful or negligent acts or omissions on the part of any service provider or its employees, (2) any defect in or failure of any vehicle, equipment or instrumentality owned, operated or otherwise used by any service provider, or (3) any wrongful or negligent acts or omissions on the part of any other party not under the control, direct or otherwise, of NMR2000.

• significant recent achievements in the theory and automation of NMR
• critical short and long term goals for NMR
• emerging new research directions in NMR
• practical applications of NMR
• significance of NMR to knowledge representation and AI in general

The program of the workshop will include special sessions on:

• Abductive reasoning
• Belief change: theory and practice
• Representing actions and planning
• System descriptions and demonstrations
• Uncertainty frameworks in NMR

We solicit papers for the workshop on all themes listed earlier as well as on other topics related to nonmonotonic reasoning and its applications.

Format requirements for regular NMR submissions are:

• 12 double-spaced pages excluding title page and bibliograpy
• on-line submissions are encouraged (postscript file)

• 6 double-spaced pages excluding title page and bibliograpy
• again, on-line submissions are encouraged (postscript file)

for an e-mail submission (preferred) send a postcript file to Mirek Truszczynski at mirek@cs.engr.uky.edu

IMPORTANT:Some special sessions have different submission requirements, submission address and submission deadlines. Please, check the appropriate announcements.

Back to the top

Important dates:

General NMR papers and posters
Special sessions on
Abductive Reasoning
Belief Revision: Theory and Practice
Planning by NMR
Uncertainty Frameworks in NMR:

Submission deadline: January 15, 2000
Acceptance decision by:  February 15, 2000
Camera ready copy due: March 8, 2000

Special session on
Implementations and System Demonstration:

Submission deadline: February 14, 2000
Acceptance decision by:  February 28, 2000
Camera ready copy due: March 8, 2000

Workshop organizers and Program Committee co-chairs:

Chitta Baral, Arizona State University (chitta@asu.edu)
Mirek Truszczynski, University of Kentucky, USA (mirek@cs.uky.edu)

Eyal Amir, Stanford University, USA
Marco Cadoli, University of Roma, Italy J
Jim Delgrande, Simon Fraser University, Canada
Marc Denecker, K. U. Leuven, Belgium
Juergen Dix, University of Koblenz, Germany
Thomas Eiter, TU Vienna, Austria
Hector Geffner, Universidad Simon Bolivar, Venezuela
Michael Gelfond, University of Texas at El Paso, USA
Fang-Zhen Lin, Hong Kong Univ of Sc and Tech
Sheila McIlraith, Stanford University, USA
Henri Prade, IRIT/CNRS, France
Torsten Schaub, University of Postdam, Germany
Hudson Turner, University of Minnesota, Duluth, USA
Mary-Anne Williams, University of Newcastle, Australia

Technical Program
 
 
 

 Back to the top

Invited talks
 
 

In this presentation we assess the present stage of the field of nonmonotonic reasoning. We discuss major developments of nonmonotonic reasoning in the last five to seven years, outline some possible new research directions and formulate several fundamental questions related to the role of nonmonotonic reasoning in knowledge representation.

While there have been many exciting developments in nonmonotonic reasoning in the last few years, in our talk we will review only some of them. First, we will discuss implementations of nonmonotonic reasoning systems such as smodels, dlv and deres, and their impact. In particular, we will talk about the emergence of the Answer-Set Programming paradigm as a consequence of demonstrated effectiveness of nonmonotonic reasoning systems. We will also discuss theoretical advances. Most notably, we will focus on the development of algebraic foundations of nonmonotonic reasoning along the lines proposed by Fitting for Logic Programming, and we will talk about the idea of program (or theory) transformations.Finally, we will look at issues of complexity, expressivity and conciseness of encodings for nonmonotonic logics.

We will also propose several intriguing research directions that are suggested by recent developments. First and foremost, we believe that in order to gain wide acceptance, the Answer-Set Programming paradigm must be integrated with mainstream programming paradigms and methodological issues of Answer-Set Programming must be studied. Given changes and advances in nonmonotonic reasoning, a major effort is needed to reexamine possible application areas of our field. Lastly, we believe that a proper treatment of quantitative issues in nonmonotonic reasoning still requires a lot of effort.

Advances in nonmonotonic reasoning give rise to several "big" questions about the role and lasting effect of nonmonotonic reasoning in knowledge representation, logic and computer science. We will look at some of these questions in the talk not so much to provide definite answers as to provoke a discussion.

Answer set programming has emerged as a viable approach to declarative logic-based knowledge representation.  It is based on the stable model semantics of logic programs and can be seen as a novel form of constraint programming where constraints are expressed as rules. The underlying idea is to encode an application problem using logic program rules so that the solutions to the problem are captured by the stable models of the rules.

What makes the approach particularly interesting is the emergence of automated reasoning tools supporting answer set programming and the rapid improvement of their performance in recent years.  This has led to the possibility of applying the paradigm to various areas including combinatorial problems, satisfiability, constraint satisfaction, planning, model checking, product configuration, and feature interaction.  The results indicate that answer set programming could provide a computationally feasible general purpose knowledge representation paradigm supporting nonmonotonic reasoning. This is a goal that has been a fundamental motivation behind the study of nonmonotonic logics since the research on nonmonotonic reasoning started in the late 1970's.

In this talk the current state of the art in answer set programming is reviewed. We start by explaining the underlying stable model semantics of logic program rules.  We discuss application methodology and review areas where answer set programming has been applied.  We examine current systems implementing the stable model semantics and consider their implementation techniques focusing, in particular, on search methods, pruning techniques, heuristics, and handling of rules with variables. We also present experimental results illustrating the current level of their performance.

In the last 20 years, nonmonotonic reasoning has established itself as an important topic of research at the core of Artificial Intelligence research. A noticeable feature of this research trend is its emphasis on a modelling of uncertainty which is not numerical. However there is no consensus as to what "non numerical" means. We first propose a typology of scales for the representation of uncertainty, ranging from purely ordinal scales (equivalently, using relations among events), to full-fledged numerical, and including  absolute scales or  scales of integers.
 

The simplest representation uses  partial or total ordering relations over a set of possible worlds.  The simplicity of this approach stems from the possibility of deriving relative levels of likelihood of propositions from a relation on possible worlds (contrary to more traditional uncertainty relations such as comparative probability). Moreover this ordinal theory of uncertainty is not self-dual (like probability) because it tells certainty from mere plausibility (in the spirit of modal logics). This simple view of uncertainty is underlying various topics such as belief revision, conditional knowledge bases in the style of Lehmann and colleagues, information fusion, and the representations of relevance.

The aim of this talk is to discuss this simple approach to uncertainty as a basis for nonmonotonic reasoning,  in connection with and opposition to other more traditional views, often numerical, such as probability theory and some extensions of variants thereof such as possibility theory. Various semantics of nonmonotonic consequence relations are recalled including a standard probabilistic one, based on"big-stepped" probabilities.

It is recalled that the uncertainty relations involved in nonmonotonic reasoning model the notion of acceptance as a deductively closed notion, contrary to probabilistic reasoning. This is due to the idea of "jumping" to plausible conclusions" that is captured by the axioms of conditional inference, as opposed to the idea of "reasoning on the average" that pervades probabilistic methods.

The notion of independence in  ordinal uncertainty theories also noticeably differs from the one in probability theory. Two notions become distinct, that could be called "decompositional independence" and "causal independence", which coincide in the probabilistic setting. The latter is closely related to belief revision, and expresses the idea that "my belief in a proposition is not affected by hearing that another proposition is true". Decompositional independence pertains to the possibility of decomposing ordering relations on Cartesian products in terms of their projections, for the sake of easier computations. It is closely related to preferential independence in measurement theory.
 

Lastly this talk will provide some material for addressing the pending question: what is the expressive power of ordinal uncertainty models as opposed to numerical ones? This question will be exemplified on two standard problems addressed by means of probability theory: causal diagnosis using prior information and likelihood functions on the one hand, and decision-making under uncertainty on the other hand. A purely ordinal approach to conditioning is less expressive than a numerical approach as to the allowed updates of prior information. Similarly, enforcing an ordinal representation of uncertainty in Savage theory of decision-making under uncertainty leads to poor and counterintuitive decision rules.

Concludingly, while simple and purely ordinal representations of uncertainty offer a very natural framework for nonmonotonic reasoning, and especially the derivation of plausible conclusions under incomplete information, their expressive power for modelling other kinds of cognitive tasks seems to be, perhaps unsurprizingly, somewhat limited. However, it is not clear that improving this expressive power implies that full-fledged numerical modelling is always unavoidable. Other "less numerical" approaches offer more expressive, although not toodemanding, frameworks.
 

This talk is based on joint works with various people including Salem Benferhat, Helene Fargier, Hector Geffner, Patrice Perny, and Henri Prade.

 Back to the top