Another to add to the list of cool looking things that I don’t have time to explore as much as I want, the Bayesian Network Tools for Java. From the bit I saw on comp.lang.java.announce:
Second, the Kansas State University Laboratory for Knowledge
Discovery in Databases is pleased to announce the initial pre-alpha
of Bayesian Network Tools in Java (BNJ) v.2.
BNJ is an open-source suite of Java tools for research and
applications development in probabilistic reasoning, and currently
includes the following modules:
- Exact inference (junction tree, variable elimination)
- Sampling-based inference (logic sampling, importance sampling)
- K2 algorithm for structure learning
- Experimental packages for structure learning
(stochastic score-based structure search, variable ordering)
- Format interconversion suite
In an alternate history of my life where I didn’t move from Kansas to Georgia at age 17, I would have gone to Kansas State. I looked at the sourceforge page and this looks like a decent enough package. It supports the XML-BIF XML Bayesian Interchange Format. I still have visions of a service on the computer that can do general classification of any text, that could be used to filter anything – e-mail, news posts, web pages, files, etc. I wonder if this gets closer to that?