-
Instructors: Ross Shachter and Prakash Shenoy
In the Foundations session, Ross Shachter and Prakash Shenoy introduced the basic principles of reasoning under uncertainty. The first part of Foundations included a presentation of important historical background, foundations of probability and decision making, and an introduction to the representation of uncertain knowledge with Bayesian networks and influence diagrams. In the second part of Foundations, Prakash Shenoy moved beyond probability theory to present alternative formalisms for reasoning under uncertainty. His discussion covered Dempster-Shafer belief functions, possibility theory, and work on abstraction of probability theory, including Spohn’s perspective on belief.
-
Instructors: Bruce D’Ambrosio, Mark Peot, and Finn Jensen
In the Inference Algorithms session, Bruce D’Ambrosio reviewed the basic principles of probabilistic inference algorithms with Bayesian networks. He covered the family of algorithms developed for inference and discussed their behaviors and applicability. Mark Peot discussed techniques for computing optimal policies in influence diagrams. Finally, Finn Jensen examined commonalities among inference algorithms in probabilistic and nonprobabilistic reasoning frameworks.
-
Instructors: Kathryn Laskey and Michael Shwe
Kathy Laskey and Michael Shwe reviewed problems and with the structuring and assessment of Bayesian networks and influence diagrams. Real-time knowledge acquisition was planned for this session so the audience could experience firsthand some of the real world issues involved with building models for reasoning under uncertainty.
-
Instructors: Greg Cooper, David Heckerman, and Wray Buntine
Wray Buntine, Greg Cooper, and David Heckerman introduced the fast growing area of learning graphical models from data. First, Greg Cooper and David Heckerman presented the foundations of learning graphical models, taking a causal perspective on influences among variables. They reviewed scores and search methods for model selection, including techniques from Bayesian statistics, neural-network research, and machine learning. After the presentation of basics, Wray Buntine described key factors to consider in the real-world application of the learning methods.
-
Instructors: Eric Horvitz and Mark Peot
Several case studies were presented that highlight multiple issues with the construction and fielding of real-world systems that rely on reasoning under uncertainty.