PublicationsFrom: Henry Tirri tirri@cs.helsinki.fiDate: Wed, 08 Nov 2000 15:35:03 +0200 Subject: B-Course: Bayesian tutorial and data analysis service B-Course is a web-based interactive tutorial on Bayesian dependence modeling. However, B-Course is more than just a tutorial: it is also a free data analysis tool that makes it possible for you to use your own data in the tutorial. Consequently, B-Course can be used as an analysis tool for any research where dependence modeling is of interest. B-Course extracts both Bayesian Belief Network models and Causal models (as discussed in Judea Pearl's recent book "Causality - Models, Reasoning and Inference") from a given data set, and supports two graphical user interfaces for inference with the constructed Bayesian networks. In more detail: B-Course allows missing data, automatically discretizes continuous variables, builds Bayesian Belief Networks, Naive Causal Models (no latent variable assumption) and Restricted Latent Variable Causal Models. The system provides both a "vanilla" user interface for Bayesian inference in the constructed Bayesian network, and a more sophisticated graphical user interface for browsers supporting Java. B-Course also supports exportation of the constructed models in Hugin Lite format, and offers a library of tutorial material about the Bayesian approach to dependency modeling. B-Course is offered free for educational and research purposes only. B-Course is a service, not a program that is downloaded. Thus it is used via a web browser and all the computationally intensive tasks are performed at the B-Course server farm. Although a considerable effort has been put in making B-Course to work with all Netscape and Explorer browser versions, it cannot be guaranteed that B-Course works with very early versions of browser software. B-Course can be found at http://b-course.cs.helsinki.fi/ B-Course Service is hosted by Complex Systems Computation Group (CoSCo), Department of Computer Science, University of Helsinki - Home of Linux. |
Copyright © 2000 KDnuggets. Subscribe to KDnuggets News!