-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-* The uncertainty list digest is about automatic reasoning under uncertainty. Measures of link strength Two books of note An SPIE session amenable to uncertainty talks. This issue contains a few things that have crossed my desk, in addition to e-mail traffic. I encourage readers also to send in items of interest to the rest of the community, such as reviews and listings of recent papers and books, conferences, both past and upcoming, and funding related items. I will try to make up the deficit with items I am aware of from other places. I have been talking to sumex about disk space for ftp access, so that we can start a library to exchange source code. Details will be forthcoming. Please use the address agosta@sumex-aim.stanford.edu for both submissions and for address change requests. -John Mark Agosta 415/965-1990 _________________________________________________________________ Date: 15 Oct 91 3:31 -0700 From: Brent Boerlage To: agosta@sumex-aim.stanford.edu Message-Id: <142*boerlage@cs.ubc.ca> Subject: Measures of link strength Hello, I am currently doing research on the notion of assigning a "strength" to BN links based on the link matrix (contingency table) of the link. The strength of the link is based on the maximum change in beliefs of the child node, for any change in beliefs of the parent. The strengths may be chained for links in series (with a special rule for nodes that are the child of two or more nodes), and global limits for the maximum change in belief in any node of a complex network may be made under some conditions. I am wondering if anyone else has worked on this problem, or knows of someone who has. Thank You, Brent Brent Boerlage boerlage@cs.ubc.ca Computer Science Dept. University of British Columbia Vancouver, B.C. V6T 1W5 _________________________________________________________________ Subject: Two books of note Thomas L. Dean & Michael P. Wellman Planning and Control (San Mateo, CA: Morgan Kaufmann, 1991) ISBN 1-55860-209-7 Phillip Smets, E.H. Mamdani, Didier Dubois, & Henri Prade Non-Standard Logics for Automated Reasoning (NY: Academic Press, 1988) Tom's book came out this summer. One might not surmise from the title that significant parts of the book cover decision-theoretic approaches. In particular, chapters 7 and 8 introduce probabilistic networks, and apply them to planning and control. More generally, the book sets out on a comprehensive synthesis of techniques from artificial intelligence, control theory, operations research and decision science. A researcher might read it for an understanding of recent probability networks work in the terms of the more traditional areas of control, and the somewhat parallel, more conventional AI approaches to planning. The "Europeans" book is the result of a pair of (ESPRIT funded?) workshops in 1986 and 1987. Each chapter is an exposition of a different topic in qualitative or quantitative logics. Chapter 8, on Probabilistic logics is written by Gerhard Paass. It appears that this book may serve as a comprehensive tutorial on various attempts at the expression on uncertainty in reasoning. __________________________________________________________________ Subject: An SPIE session amenable to uncertainty talks. I have included this clipping from the San Diego SPIE conference schedule, to point to a new forum for exchange on uncertainty and recognition approaches. The conference chair, Su-Shing Chen, explained that he hopes to attract a spectrum of uncertainty approaches to image recognition. He plans to run a similar session in coming years. He will be moving to NSF next year. ------ SPIE's 1991 International Symposium on Optical and Applied Science and Engineering: Conference 1569, 24 July 1991. Stochastic and Neural Methods in Signal Processing, Image Processing, and Computer Vision conference Chair: Su-Shing Chen, U. of NC Session 1: Evidential Reasoning and Belief Networks chair: Seth Hutchinson, U of IL Using Bayesian networks to combine evidence from multiple sources for image segmentation. S. M. LaValle, S. A. Hutchinson, U. of IL An Example of a Bayes Network of relations among visual features J. M. Agosta, Stanford U. Recursive computation of a wire-frame representation of a scene form dynamic stereo using belief functions. A. P. Tirumalai, B. G. Schunck, R. C. Jain, U. of MI Evidential reasoning in the PSEIKI image interpretation. K. M. Andress, Siemens Application of Dempster-Shafer theory to a novel control scheme for sensor fusion. R. R. Murphy GA Inst of Tech. (end of the uncertainty list, 18 Oct 1991)