Solomon Eyal Shimony

Associate Professor


B.S.E.E.: Technion, I.I.T. - 1982

Ph.D.: Brown University - 1991

New (October 2009): "job" offers

Need new graduate students (both Masters and PhD candidate levels) in AI: decision-making under uncertainty, and learning. For IMG4 consortium and other funded research projects.

Research Interests and Agenda

My main research area is artificial intelligence, with a focus on uncertain reasoning and its applications. When considering uncertain reasoning, Bayesian probabilistic reasoning has several advantages over other uncertainty formalisms, when applied to the real world: 1) It is relatively well understood, and can be integrated within a decision-theoretic sensing and acting. 2) The well understood tool of estimation theory is also based on probabilistic semantics. 3) The required distributions can be learned by experiment. Uncertainty is represented as a distribution over a sample space, and can structured by using graph models, such as directed acyclic Bayesian Belief Networks (Bayes nets).

In earlier work, I provided semantics for weighted abductive reasoning, and incorporated irrelevance into probabilistic abduction, in the framework of Bayes nets. The complexity of probabilistic reasoning and problems derived from it have been the focus of several of my publications. These include complexity of belief revision, tractable problems and algorithms for weighted proof graphs (WAODAGs), and approximate belief updating. Belief updating is a hard problem, but approximation algorithms for belief updating and revision on Bayes nets seem to be useful in practice.

One of our approximation algorithms works by enumerating high-probability instantiations to the network variables. Another uses sampling of partial instantiations. In a newer venture, various algorithms are combined within a decision-theoretic meta-reasoning framework, to take advantage of the better characteristics of the different algorithms. Such schemes are becoming increasingly popular under the framework of flexible computation, that allows trading off resources for quality of results. Another of my related research topics are in applying uncertain reasoning to constraint satisfaction problems (CSP).

For the present and the near future, my agenda includes looking into more refined models, such as Bayesian Knowledge Bases, which are more general than Bayes nets in that they allow cycles in the directed graph. Additionally, application areas such as sensor fusion for robotics, and using probabilistic networks in data-mining seem to me to be fruitful ground for future development.

Professional Activities

  • Co-organizer, Israeli Association for Artificial Intelligence symposium, 2008.
  • Chair, Paul Ivanier Center for Robotics and Production Management, 2004-2008.
  • EX-head of teaching committee for the undergraduate Computer Science program
  • Program co-chair, IEEE International Conference on Systems, Man, and Cybernetics 2006.
  • BISFAI-99 program co-chair - guest editor of special issue for selected papers - Annals of Math. and AI, Applied Intelligence.
  • Reviewer, JAIR, Artificial Intelligence, Approximate Reasoning, IEEE SMC, Annals of Math. and AI
  • Program Committee member, Uncertainty in Artificial Intelligence, 1994, 1995, 1997-2006

    Courses taught in past

    Contact Addresses

    e-mail shimony@cs.bgu.ac.il

    Physical Address 

    Building 37 (Alon Hi-Tech Building)
Room 216
Office hours: Wednesday 14-16

    Snail-Mail Address:

        Department of Computer Science
    Ben-Gurion  University of the Negev
    P.O. Box 653
    84105 Beer-Sheva
    Israel

    Tel: (+972-8) 647-7857
    FAX: (+972-8) 647-2909