Uncertainty Digest Tuesday, April 14th 1992 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ ~ The uncertainty list a means for researchers in uncertain reasoning ~ to exchange information, such as conference announcements, articles, ~ reviews, and software. ~ ~ ~ Please send all submissions and correspondence, as well as requests ~ for changes to the mailing list to agosta@sumex-aim.stanford.edu. ~ Archives may be found by anonymous ftp from sumex-aim.stanford.edu ~ in /var/ftp/pub/unc ~ ~ ~ This list is maintained by John Mark Agosta, 415/859-4931 ~ My apologies to the community for the two month delay in transmission. ~ I have recently started at SRI International, and have spent much ~ time getting re-organized. Private mail may be sent to ~ johnmark@sri.com ~ ~ ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ Today's Topics: Special Issue of the J. of AI in Medicine (deadline June 1) KR'92 - CALL FOR PAPERS (deadline Apr 21) The Third International Workshop on Principles of Diagnosis (deadline June2) ftping a demo of HUGIN, a Bayesian Belief Network Shell Workshop on interval methods in AI (deadline Aug 1) ___________________________________________________________________________ Special Issue on Probabilistic and Decision-theoretic Systems in Medicine in the journal Artificial Intelligence in Medicine This is a reminder that papers are sought for a special issue of the journal Artificial Intelligence in Medicine on the use of probability and decision theory to develop medical decision support systems. Papers will be refereed for their significance, originality, quality, and clarity. Three copies of a submitted paper must arrive at the address below by June 1, 1992. The author(s) should attach the following information to each copy of the paper: postal address, email address, telephone number, and fax number. Please address all correspondence and send papers to the special issue editor at the following address: Greg Cooper B50A Lothrop Hall 190 Lothrop Street University of Pittsburgh Pittsburgh, PA 15261 e-mail: gfc@med.pitt.edu phone: (412) 648-3190 fax: (412) 648-3126 ______________________________________________________________________ KR'92 - CALL FOR PAPERS THIRD INTERNATIONAL CONFERENCE ON PRINCIPLES OF KNOWLEDGE REPRESENTATION AND REASONING Royal Sonesta Hotel, Cambridge, Massachusetts, USA with support from AAAI, ECCAI, and CSCSI in cooperation with IJCAII October 26-29, 1992 (KR'92 follows the AAAI Fall Symposium Series at the same location October 23-25) The idea of explicit representations of knowledge manipulated by inference algorithms provides an important foundation for much work in Artificial Intelligence, from natural language to expert systems. A growing number of researchers are interested in the principles governing systems based on this idea. This conference will bring together these researchers in a more intimate setting than that of the general AI conferences. In particular, authors will have the opportunity to give presentations of adequate length to present substantial results. The theme of this year's conference is the relationship between the principles of knowledge representation and reasoning and their embodiment in working systems. Authors are encouraged to relate their work to one of the following important questions: (1) What issues arise in applying knowledge representation systems to real problems, and how can they be addressed? (2) What are the theoretical principles in knowledge representation and reasoning? (3) How can these principles be embodied in knowledge representation systems? Submissions are encouraged in (but are not limited to) the following topic areas: KNOWLEDGE REPRESENTATION FORMALISMS REASONING METHODS - logics of knowledge and belief - deduction - nonmonotonic logics - abduction - temporal logics - induction - spatial logics - learning - taxonomic logics - planning and plan analysis - logics of uncertainty - constraint solving and evidence - diagnosis - classification - inheritance - belief management and revision - analogical reasoning GENERIC ONTOLOGIES FOR DESCRIBING ISSUES IN IMPLEMENTED KR&R SYSTEMS - time - comparative evaluation - space - empirical results - causality - benchmarking and testing - resources - reasoning architectures - constraints - efficiency/completeness tradeoffs - applications classes - complexity such as medicine - algorithms SUBMISSION OF PAPERS The Program Committee will review EXTENDED ABSTRACTS rather than complete papers. Abstracts must be at most twelve (12) pages with a maximum of 38 lines per page and an average of 75 characters per line (corresponding to the LaTeX article-style, 12pt), excluding the title page and the bibliography. Overlength submissions will be returned. All abstracts must be submitted on 8 1/2" x 11" or A4 paper, and printed or typed in 12-point font (10 characters/inch on a typewriter). Dot matrix printout, FAX, or electronic submission will not be accepted. Each submission should include the names and complete addresses of all authors. Correspondence will be sent to the first author, unless otherwise indicated. Also, authors should indicate under the title which of the questions and/or topic areas listed above best describes their paper (if none is appropriate, please give a set of keywords that best describe the topic of the paper). Abstracts must be received by one of the program co-chairs no later than April 21, 1992. Papers received after that date will be returned unopened. Authors will be notified of the Program Committee's decision by June 15, 1992. REVIEW OF PAPERS Submissions will be judged on clarity, significance, and originality. An important criterion for acceptance is that the paper clearly contributes to principles of representation and reasoning that are likely to influence current and future AI practice. Extended abstracts should contain enough information to enable the Program Committee to identify and evaluate the principal contribution of the research and its importance. It should also be clear from the extended abstract how the work compares to related work in the field. Submitted papers must be unpublished. Submissions must also be substantively different from papers currently under review and must not be submitted elsewhere before the author notification date (June 15, 1992). FINAL PAPERS Authors of accepted papers will be expected to submit substantially longer full papers for the conference proceedings. Final camera-ready copies of the full papers will be due August 3, 1992. Final papers will be allowed at most twelve (12) double-column pages in the conference proceedings (corresponding to approx. 28 article-style LaTeX pages; a style file will be provided by the publisher). CONFERENCE CHAIR Charles Rich Mitsubishi Electric Research Laboratories 201 Broadway Cambridge, MA 02139, USA Voice: +1 (617) 621-7507 Fax: +1 (617) 621-7550 Email: rich@merl.com PROGRAM CO-CHAIRS Bernhard Nebel William Swartout DFKI USC/Information Sciences Institute Stuhlsatzenhausweg 3 4676 Admiralty Way D-W-6600 Saarbrucken Marina del Rey, CA 90292-6695 Germany USA Voice: +49 (681) 302-5254 Voice: +1 (213) 822-1511 Fax: +49 (681) 302-5341 Fax: +1 (213) 823-6714 Email: nebel@dfki.uni-sb.de Email: swartout@isi.edu LOCAL ARRANGEMENT CHAIR James Schmolze Dept.of Computer Science Tufts University Medford, MA 02155 USA Voice: +1 (617) 627-3681 Fax: +1 (617) 627-3443 Email: schmolze@cs.tufts.edu PROGRAM COMMITTEE James Allen (Univ of Rochester), Guiseppe Attardi (Univ of Pisa), Daniel Bobrow (Xerox PARC), Ron Brachman (AT&T Bell Labs), Gerd Brewka (GMD, Bonn), Rina Dechter (UC Irvine), Johan de Kleer (Xerox PARC), Jon Doyle (MIT), David Etherington (AT&T Bell Labs), Richard Fikes (Stanford Univ), Alan Frisch (Univ of Illinois), Dov Gabbay (Imperial College), Michael Georgeff (AAII), Pat Hayes (Stanford Univ), Maurizio Lenzerini (Univ of Roma), Robert MacGregor (USC/ISI), Alan Mackworth (UBC), David Makinson (Paris), David McAllester (MIT), Fumio Mizoguchi (Science Univ of Tokyo), Wolfgang Nejdl (TU Vienna), Hans-Juergen Ohlbach (MPI, Saarbruecken), Peter Patel-Schneider (AT&T Bell Labs), Ramesh Patil (USC/ISI), Judea Pearl (UCLA), Martha Pollack (Univ of Pittsburgh), Henri Prade (Univ Paul Sabatier), Erik Sandewall (Univ of Linkoeping), Len Schubert (Univ of Rochester), Stu Shapiro (SUNY Buffalo), Gert Smolka (Univ of Saarland, DFKI Saarbruecken), Peter Szolovits (MIT), Mike Wellman (USAF Wright Lab) IMPORTANT DATES Submission receipt deadline: April 21, 1992 Author notification date: June 15, 1992 Camera-ready copy due to publisher: August 3, 1992 Conference: October 26-29, 1992 _____________________________________________________________________ The Third International Workshop on Principles of Diagnosis (DX-92) Seattle, Washington, USA, or vicinity October 11-14, 1992 Call for Papers This is the third in a series of workshops to encourage interaction and cooperation among researchers in artificial intelligence with diverse approaches to diagnosis. Previous workshops in this series were held at CISE, Milano (Italy) in 1991 and at Stanford University (USA) in 1990, with a forerunner at Paris (France) in 1989. Attendance will be limited to fifty participants with three days of presentations and substantial time reserved for discussion. Those interested in participating should submit papers for review by the committee. Submissions are welcome on (but not limited to) the following topics: o Theory of diagnosis: abductive, deductive, or probabilistic theories. o Computational issues: controlling combinatorial explosion; focusing strategies; controlling inference in complex systems; use, inference, or absence of structural knowledge. o Modeling for diagnosis: multiple, approximate, incomplete, probabilistic, and qualitative models; integration of heuristics with model-based diagnosis; geometric knowledge; dynamic systems; embedded software. o Integration of diagnosis with control, planning, analysis, theory construction, tutoring, and design/redesign, testability/ diagnosability/sensor-placement. o Empirical evaluation of theoretical results, theoretical implications of practical applications and their successes or failures. o Inductive approaches to diagnosis: case-based reasoning, neural nets. Although not a requirement, previously unpublished work is preferred. Papers are limited to a maximum of 5000 words; shorter papers are encouraged, but space should be used to ensure adequate presentation. Include postal (and courier) addresses, electronic mail, fax, and telephone numbers. Please indicate whether you wish to present or only attend. The conference chair (below) must receive three paper copies of each submission by June 2, 1992, and notifications will be sent by August 4. Accepted papers can be revised for inclusion in the workshop working notes. Workshop chair: Ethan Scarl; postal address (NOT for courier deliveries): MS 7L-64, Boeing Computer Services, P.O. Box 24346, Seattle, WA USA 98124-0346; courier address (NOT for US mail): MS 7L-64, Boeing Computer Services 2760 - 160th Avenue S.E., Bldg. 33-07, Bellevue, WA USA 98008; phone: (206) 865-3255; fax: (206) 865-2964; email: ethan@atc.boeing.com. Committee: I. Bratko (U. Ljubljana), L. Console (U. di Udine), P. Dague (IBM), J. deKleer (Xerox), O. Dressler (Siemens), K. Eshghi (HP, Stanford), G. Friedrich (T.U. Wien), D. Heckerman (USC), M. Kramer (MIT), W. Nejdl (T.U. Wien), R. Patil (ISI), D. Poole (U. British Columbia), O. Raiman (Xerox), M. Shirley (Xerox), J. Sticklen (Michigan State U.), G. Tornielli (CISE), M. Wellman (Wright-Patterson), B. Williams (Xerox) ______________________________________________________________________ HUGIN, a Bayesian Belief Network Shell. ------------------------------------- HUGIN is a graphical environment for constructing, testing and experimenting with knowledge bases represented by Bayesian belief networks. The graphical user-interface of HUGIN is developed in X-windows and allows for interactive creation and maintenance of knowledge bases. A compiler transforms the Bayesian belief network into an computational efficient structure. The compiler offers a variety of options in order to generate the most efficient structure given the domain at hand, such as efficient triangulation methods, approximation- and compression capabilities. A runtime systems provides facilities for easy entering and propagation of information. The HUGIN software is written in ANSI-standard C and provides an application programmer's interface (API). The API is a collection of C functions which provides access to and control of the compilation of Bayesian belief networks. It includes functions for entering and retrieving data form domain knowledge bases and the necessary functionality to perform reasoning. The API can be used in C programs as well as from programs written in any language capable of calling external C functions. This renders a smooth way of integrating the HUGIN inference engine in complex systems. The HUGIN software represents state of the art in Bayesian belief network software and has already been proven to be a robust, reliable environment for building large Bayesian belief networks. How to get a demo-copy: --------------------- The HUGIN system version 2.0 is a commercial product, available at corporate and academic prices. A demonstration version of HUGIN for the SPARC platform is available for anonymous ftp from the following sites: Europe (Denmark) Site: hugin.dk (129.142.51.10) Directory: /pub File: hugin.tar.Z North America (California) Site: sumex-aim.stanford.edu (36.44.0.6) Directory: /var/ftp/pub/unc/solvers/hugin File: 1800 -rw-rw-rw- 1 agosta 1829597 Apr 5 16:01 hugin.tar.Z Please follow the directions in the README file. How to get more information: ---------------------------- For further details or questions please contact Hugin Expert or our agent in North America, Knowledge Industries Inc. Electronic Mail (Internet): info@hugin.dk Hugin Expert Ltd. Phone: +45 9815 6644 Fax : +45 9815 Surface Mail: Hugin Expert Ltd. Niels Jernes vej 10 DK-9220 Aalborg O Denmark Agent in North America: Knowledge Industries Inc. Phone: 415 321 9521 Fax : 415 326 1769 Surface Mail Knowledge Industries 125 California Avenue Palo Alto, CA 94306 USA ______________________________________________________________ Workshop on interval methods in artificial intelligence February 26 through March 1, 1993, Lafayette, Louisiana in conjunction with the conference on NUMERICAL ANALYSIS WITH AUTOMATIC RESULT VERIFICATION WHY INTERVALS IN AI? Main goals of Artificial Intelligence (AI) research are to represent human knowledge (both common sense and expert knowledge) and to process it (i.e. to enable computers to do thereasoning). Human knowledge is usually not absolutely precise. For example, a phrase like ``the temperature is moderate today'' does not give a precise value of the temperature, but rather an interval of possible values. In many other cases natural formalization of human knowledge leads to intervals: in reasoning about time and actions, in common sense reasoning about physical quantities, etc. Similar methods are used to represent human knowledge about 2- and 3-dimensional images: e.g., in scene analysis, in expert systems for navigation, for controlling themotion of robots, etc. Different statements constituting human knowledge have different degrees of certainty: we are more certain in some of themthan in some others. So in addition to representing these statements,one should somehow represent their degrees of uncertainty. In traditional mathematics, probability theory is a well- developed formalism in which probabilities represent such degrees of uncertainty.If we have complete knowledge of the frequencies with which different statements turn out to be true, we can get precise values for the corresponding probabilities. But in AI such complete information is usually lacking, so instead of a single probability we can characterizethe uncertainty by an interval of possible probability values. Such a characterization is explicit, for example, in Dempster-Shafer formalism. Similar intervals occur naturally in fuzzy logic and the certainty values approach. INTERVAL MATHEMATICS CAN BE HELPFUL. Since information is represented via intervals, processing this information involves processing interval data. AI researchers have invented numerous new methods for that purpose. However, Interval Analysis is an established area of applied mathematics which deals precisely with processing intervals. The conference as a whole is devoted to various aspects of interval analysis. Within this workshop, held in conjunction with the conference, interaction between AI researchers and researchers in interval analysis can lead to the effective application of interval methods to the solution of AI problems. THE WORKSHOP'S MAIN PURPOSE IS CROSS FERTILIZATION. We want to bring researchers in interval analysis and researchers who apply interval methods in AI together. AI researchers should benefit from a better knowledge of the state of the art in interval processing. Specialists in interval analysis will learn ofmany new challenging problems of great practical and philosophical importance, to which they can apply their methods. TOPICS. The workshop will cover all possible applications of interval methods in AI, including (but not limited to): - intervals in reasoning about time and actions - intervals in common sense reasoning about physical quantities - intervals and similar methods in representing 2- and 3- dimensional images - intervals in uncertainty representation and uncertainty reasoning (including intervals appearing in Dempster-Shafer theory, fuzzy reasoning, etc). PROGRAM COMMITTEE: Chitta Baral (University of Texas at El Paso) Bart Kosko (University of Southern California) Vladik Kreinovich (University of Texas at El Paso) Raymond Ng (University of Maryland) V. S. Subrahmanian (University of Maryland) Patrick Suppes (Stanford University) Lotfi Zadeh (University of California at Berkeley) (others may possibly be included later) CALL FOR PAPERS. Papers on all aspects of interval processing in AI are welcome. Contributors should send a one page abstract to Secretary of the Program Committee Vladik Kreinovich Computer Science Department University of Texas at El Paso Office phone: (915) 747-5470 Fax: (915) 747-5616 email: vladik@cs.ep.utexas.edu The abstract may be sent electronically if it is in an ASCII format or in some version of TeX (common versions of TeX are preferable). The deadline for abstracts is August 1, 1992. We are planning to publish refereed proceedings. CONFERENCE LOCATION, REGISTRATION FEE, LOCAL ARRANGEMENTS: see attached information about the main conference on Numerical Analysis with Automatic Result Verification. APPENDIX: REGISTRATION FORMS and FURTHER INFORMATION for an International Conference on NUMERICAL ANALYSIS WITH AUTOMATIC RESULT VERIFICATION: Mathematics, Applications, and Software [ For this appendix, etc. and registration, see the file 9 -rw-r--r-- 1 agosta 9164 Apr 8 18:29 intervals.annoc in sumex-aim:/var/pub/ftp/unc/announcements - jma] __________________________________________________________________ *** end of uncertainty digest