Public Domain Belief Network Tools:
GeNIe: Nice graphical interface for defining belief networks with discrete variables. Built on top of SMILE, a library of representation and inference algorithms that you can embed into your applications. UPitt also has made available a library of belief networks and influence diagrams drawn from textbooks and the literature.
WinBUGS: BUGS is a software package for inferring the parameters of statistical models expressed as belief networks. This is probably the premier model for learning the parameters of a belief network from data. WinBUGS is BUGS with a slick Windows interface.
Hugin: The graphical interface is less nice, but the underlying algorithms are powerful. Allows discrete (multinomial) and continuous (gaussian) variables and can reason with mixture distributions (weighted sums of gaussians). The full version of the program is expensive, but there is a "lite" version that is available for free.
MSBN32:This is the tool that Microsoft uses to prototype several of its knowledge bases including Microsoft’s online troubleshooters (http://support.microsoft.com/support/tshoot/) and the Home Health Pregnancy and Childcare system. Available for noncommercial use from Microsoft.
JavaBayes:A Java applet that you can use to construct and evaluate belief networks. Presumably works under any platform that supports Java.
Netica:Norsys develops nice tool for drawing and evaluating both belief networks and influence diagrams. The site has a large library of examples that you can play with. The tool has a programmer’s API. Available for the PowerPC, 68K Macintosh, or Windows 95/98/NT.
Analytica: Lumina Decision Systems has developed the premier tool for developing predictive models. This tool is used by lots of folks for developing belief networks for modeling environmental problems, business decisions, public policy decisions, etc. This tool has won several awards for decision support software. Analytica is available for both Windows and Macintosh platforms. A limited time trial version is available for download from
Autoclass: Autoclass is a Bayesian software package for finding clusters in data.
Other Interesting Sites:
The Air Force Institute of Technology maintains a large and useful belief network site, including software downloads, a bayes net tutorial and a literature review:
The Association for Uncertainty in Artificial Intelligence has sponsored a conference on uncertain reasoning techniques every year since 1985. In recent years, this conference has focussed almost exclusively on probabilistic reasoning. The highlight of this site is an online collection of the papers from UAI-94 through UAI-98.
Copyright 1999, Mark Alan Peot