We are pleased to announce GeNIe 1.0, a free development
environment for graphical decision-theoretic models. At the
moment it implements Bayesian networks and influence diagrams
with, what we believe to be, a pleasant and reliable user interface.
It has hierarchical submodels, a Windows-style tree view,
Noisy-OR nodes, deterministic nodes, multiple decision nodes,
relevance reasoning that includes designating nodes as targets,
multiple utility nodes, linearly-additive Multi-Attribute
utility nodes (we will have generalized MAU nodes in the future),
value of information computation, several Bayesian networks
algorithms to choose from, on-screen comments, HTML-based help
system, and many other useful features that one would want from
a development environment for graphical models.
GeNIe is our research and teaching vehicle, so naturally
it will evolve as time goes.
One feature that may contribute to sharing research results in
the UAI community is that GeNIe is able to read and write several
file formats, including old Ergo format, Netica, Hugin, and the
version of the interchange format implemented by Microsoft's MSBN
package. It can thus be used as a conversion program.
GeNIe runs on Windows 95/NT computers. We will soon release its
portable inference engine, DSL_GNL, with C/C++ inferface. DSL_GNL
can be embedded in applications that use models developed using
GeNIe. Stay tuned!
GeNIe is free for the taking (for non-commercial use, such as
basic research, teaching, and personal use). You can download
it from the following location:
http://www2.sis.pitt.edu/~genie
We believe that the program will prove useful for the community.
May GeNIe grant you your three wishes :-).
Marek Druzdzel