Below we provide a selection of web-sites with useful information and software concerning probabilistic networks and graphical models, taken from the book
R. G. Cowell, A. P. Dawid, S. L. Lauritzen and D. J. Spiegelhalter (1999).
Probabilistic Networks and Expert Systems. Springer-Verlag, New York.
A good starting point is the
Association for Uncertainty in Artificial Intelligence
site. Clearly no recommendation or guarantee can be given as to the
accuracy or suitability of any of the software mentioned.
Russell Almond's page on software for manipulating belief networks:
Contains an extensive, but not necessarily up-to-date, list of both non-commercial and commercial software, with a glossary and reference list.
Kevin Murphy's page with an introduction to Bayesian networks:
Contains a review, a good list of recommended reading with links to downloadable versions, and a page of free Bayesian network software (see below).
Contains links to proceedings of past Uncertainty in Artificial Intelligence conferences, as well as testimonials on the use of Bayesian networks, downloadable tutorials, and an excellent list of links to related sites of organizations, companies and individuals.
Decision Support Systems Group at Aalborg University:
Describes their work on methodological development and practical application of Bayesian networks and influence diagrams.
Describes the background and aims of this initiative of the European Science Foundation, which brought together researchers who use stochastic models that exploit conditional independence. It puts research on probabilistic expert systems into a broader context including Bayesian computation, genetics, environmental modelling and image analysis.
Microsoft Research Decision Theory and Adaptive Systems Group
Gives brief description of projects being undertaken in information retrieval, diagnostics and troubleshooting, intelligent user interfaces and so on, and how this work is being incorporated into Microsoft products. Some publications, particularly on learning models from data, are downloadable. Their free software, Microsoft Belief Networks MSBN is available (see next section).
Just a few links are given here --- see the above pages for a fuller and more up-to-date guide.
Russell Almond's page on software for learning belief networks from data:
Contains an extensive, but not necessarily up-to-date, list of both non-commercial and commercial software, in an area which carries over into the machine learning and data-mining literature.
Provides an overview and download information for the TETRAD software for building causal models from statistical data, and provides a publication list of project participants.
Kevin Murphy's page of free Bayesian network software:
Gives links to a wide range of free packages, including his own Bayes Net Toolbox, and at the time of writing it is up-to-date.
Includes the free Bayesian Knowledge Discover program for automatically constructing discrete Bayesian networks from databases.
Has a link to the freeware program XBAIES for building chain graph probabilistic and decision networks, available for several platforms.
Fabio Cozman's free JavaBayes program:
Includes download information and program details, with further links on Bayesian networks.
Jens Henrik Badsberg's page for his CoCo software:
Describes the software for graphical modelling of
discrete variables, but is also a base for links and
distribution of various other graphical modelling
programs, including MIM, DIGRAM, BIFROST, and GAMES.
Home page of the BUGS project:
Includes the free WinBUGS program
to build Bayesian graphical models. All inferences are by Markov Chain
Monte Carlo methods.
A freeware program, developed by the Decisions System Laboratory at the University of Pittsburgh, for inference in Bayesian networks by a variety of methods. It can read and save files in most of the formats used by other similar programs. The Decisions System Laboratory also distribute a set of C++ classes, called SMILE, which programmers can use to build their own Bayesian network programs.
Commercial Bayesian network software includes:
These pages contain numerous links to sites of interest.
This page contains a bibliography and pointers to papers concerned with perfect sampling from Markov chains.