Full-Day Course: Details

Wednesday, July 31, 1996

Reed College

Portland, Oregon, USA

Twelfth Conference on Uncertainty in Artificial Intelligence

This one-day course on principles and applications of uncertain reasoning will be given on Wednesday, July 31 (the day before the start of the main UAI 96 conference) at Reed College. Registration information for the course will be available shortly.

Introduction and Goals

Eric Horvitz and Finn Jensen

Session I. Foundations of Uncertainty: 8:35-11:00am

In the Foundations session, Ross Shachter and Prakash Shenoy will introduce the basic principles of reasoning under uncertainty. The first part of Foundations will include 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 will move beyond probability theory to present alternative formalisms for reasoning under uncertainty. His discussion will cover Dempster-Shafer belief functions, possibility theory, and work on abstraction of probability theory, including Spohn's perspective on belief.

Session II. Inference Algorithms for Belief and Action: 11:00-12:45

In the Inference Algorithms session, Bruce D'Ambrosio will review the basic principles of probablistic inference algorithms with Bayesian networks. He will cover the family of algorithms developed for inference and will discuss their behaviors and applicability. Mark Peot will discuss techniques for computing optimal policies in influence diagrams. Finally, Finn Jensen will examine commonalities among inference algorithms in probabilistic and nonprobabilistic reasoning frameworks.

Lunch Break 12:45-2:00pm

Session III. Modeling and Knowledge Acquisition: 2:00-3:15

Instructors: Kathryn Laskey and Michael Shwe

Kathy Laskey and Michael Shwe will review problems and methods with the structuring and assessment of Bayesian networks and influence diagrams. Real-time knowledge acquisition is being planned for this session so the audience can experience firsthand some of the real world issues involved with building models for reasoning under uncertainty.

Session IV. Learning Models from Data: 3:15-4:20

Wray Buntine, Greg Cooper, and David Heckerman will introduce the fast growing area of learning graphical models from data. First, Greg Cooper and David Heckerman will present the foundations of learning graphical models, taking a causal perspective on influences among variables. They will review 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 will describe key factors to consider in the real-world application of the learning methods.


Session V. Uncertain Reasoning in the Real World--Case Studies: 4:30-5:25

Several case studies will be presented that highlight multiple issues with the construction and fielding of real-world systems that rely on reasoning under uncertainty.

Instructors: Eric Horvitz and Mark Peot

Research Directions / UAI 96 Highlights

Back to UAI-96 Homepage

If you have questions or comments about the UAI-96 Full-Day Course, contact the UAI-96 Program Cochairs: Eric Horvitz and Finn Jensen. For conference arrangements information, please contact Steve Hanks.