$1990 per person
This 3-day course introduces Bayesian networks (BNs) and the use of Netica to implement BNs. The basic building blocks of these graphical probability models are presented: nodes and arcs and conditional probability tables, probability theory, and causal considerations. Insight into sophisticated BN models is provided through examples and additional considerations such as D-separation and sensitivity to findings. Building BNs using both elicitation and data is covered, with some discussion for integrating data and expert judgment into a single model.
Advanced topics include (1) using the Netica API to create implementations that integrate with other software products, including Excel and ExtendSim; (2) GeoNetica for building a BN across layers of a geospatial database; and (3) dynamic Bayesian networks for explicating representing time.
Several hours of the 3-day program devoted to helping participants build their own models on topics of their choosing. The course is valuable to program analysts, engineers and scientists who are entering the field or interesting in exploring this topic. A comprehensive set of notes and references will be provided to all attendees. Students will also receive sample models and examples of code that they can use to quickly assess implement their own ideas.
What you will learn:
- The fundamental theory of Bayesian networks.
- How to build BNs in Netica.
- Critical considerations in connecting nodes.
- Elicitation methods needed for success.
- Basics in learning models from data.
- Basics in using the Netica API.
- Basics in using GeoNetica
- IntroductionA simple Bayesian network (BN) for drug testing is described and participants build a portion of it using Netica. An overview of probability brings everyone up to the level needed. A complete diagnostic BN – Liver diagnosis case study is presented.
- Elicitation of BN structure and probabilities. Methods are presented for successful elicitation of the structure and probabilities.
- Causal Models. An overview of current thinking on causal models is presented with numerous examples.
- Troubleshooting systems. Two examples of troubleshooting systems (car and industrial system) are presented in detail.
- Learning from data. Theory and software for learning both structure and probabilities from data is presented with examples.
- Using the Netica API. Examples of using the Netica API to integrate with other software to create a user interface or extend computation are presented.
- Using the GeoNetica capability. GeoNetica enables the user to build a BN for each pixel in a geospatial database using map layers and other data. Examples are presented.
- . Throughout the course participants will be given the opportunity to build their BN with help from the instructors.
Tuition for this three-day course is $1990 at one of our scheduled public courses. Onsite pricing is available. Please call us at 410-956-8805 or send an email to firstname.lastname@example.org.
Dennis Buede, Ph.D., has building and using BNs for over 30 years. He has conducted successful research with BN applications and taught this material at the graduate level at George Mason University and Stevens Institute of Technology. He has built over a hundred BNs using elicitation techniques, and nearly a dozen using data.
Dave Brown, Ph.D., did his Ph.D. research on integrating BNs with other modeling techniques. He has used this research for nearly a dozen successful modeling efforts for defense and other federal government agencies.
Bob Powers, Ph.D., has been applying BNs to geospatial problems for nearly twenty years. He has built several GeoNetica models and provided many suggestions for extending the software’s capability.