Thirteenth Conference on Uncertainty in Artificial Intelligence

August 1-3, 1997
Brown University
Providence, Rhode Island, USA

UAI '97 Conference Program

Thursday, July 31, 1997

Conference and Course Registration 8:00-8:30am

Full-Day Course on Uncertain Reasoning 8:30-6:00pm

Friday, August 1, 1997

Main Conference Registration 8:00-8:25am

Opening Remarks
Dan Geiger and Prakash P. Shenoy

Invited talk I: Local Computation Algorithms
Steffen L. Lauritzen

Abstract: Inference in probabilistic expert systems has been made possible through the development of efficient algorithms that in one way or another involve message passing between local entities arranged to form a junction tree. Many of these algorithms have a common structure which can be partly formalized in abstract axioms with an algebraic flavor. However, the existing abstract frameworks do not fully capture all interesting cases of such local computation algorithms. The lecture will describe the basic elements of the algorithms, give examples of interesting local computations that are covered by current abstract frameworks, and also examples of interesting computations that are not, with a view towards reaching a fuller exploitation of the potential in these ideas.

Invited talk II: Coding Theory and Probability Propagation in Loopy Bayesian Networks
Robert J. McEliece

Abstract: In 1993 a group coding researchers in France devised, as part of their astonishing "turbo code" breakthrough, a remarkable iterative decoding algorithm. This algorithm can be viewed as an inference algorithm on a Bayesian network, but (a) it is approximate, not exact, and (b) it violates a sacred assumption in Bayesian analysis, viz., that the network should have no loops. Indeed, it is accurate to say that the turbo decoding algorithm is functionally equivalent to Pearl's algorithm applied to a certain directed bipartite graph in which the messages circulate around indefinitely, until either convergence is reached, or (more realistically) for a fixed number of cycles. With hindsight, it is possible to trace a continuous chain of "loopy" belief propagation algorithms within the coding community beginning in 1962 (with Gallager's iterative decoding algorithm for low density parity check codes), continued in 1981 by Tanner and much more recently (1995-1996) by Wiberg and MacKay-Neal. In this talk I'd like to challenge the UAI community to reassess the conventional wisdom that probability propagation only works in trees, since the coding community has now accumulated considerable experimental evidence that in some cases at least, "loopy" belief propagation works, at least approximately. Along the way, I'll do my best to bring the AI audience up to speed on the latest developments in coding. My emphasis will be on convolutional codes, since they are the building blocks for turbo-codes. I will mention that two of the most important (pre-turbo) decoding algorithms, viz. Viterbi (1967) and BCJR (1974) can be stated in orthodox Bayesian network terms. BCJR, for example, is an anticipation of Pearls' algorithm on a special kind of tree, and Viterbi's algorithm gives a solution to the "most probable explanation" problem on the same structure. Thus coding theorists and AI people have been working on, and solving, similar problems for a long time. It would be nice if they became more aware of each other's work.

Break 10:30-11:00am

Plenary Session I: Modeling

Lunch 12:00-1:30pm

Plenary Session II: Learning & Clustering

Poster Session I: Overview Presentations

Poster Session I

Saturday, August 2, 1997

Invited talk III: Genetic Linkage Analysis
Alejandro A. Schaffer

Abstract: Genetic linkage analysis is a collection of statistical techniques used to infer the approximate chromosomal location of disease susceptibility genes using family tree data. Among the widely publicized linkage discoveries in 1996 were the approximate locations of genes conferring susceptibility to Parkinson's disease, prostate cancer, Crohn's disease, and adult-onset diabetes. Most linkage analysis methods are based on maximum likelihood estimation. Parametric linkage analysis methods use probabilistic inference on Bayesian networks, which is also used in the UAI community. I will give a self-contained overview of the genetics, statistics, algorithms, and software used in real linkage analysis studies.

Plenary Session III: Markov Decision Processes

Break 10:30-11:00am

Panel Discussion: 11:00-12:00am

Lunch 12:00-1:30pm

Plenary Session IV: Foundations

Poster Session II: Overview Presentations

Poster Session II

Saturday Evening, August 2, 1997

UAI '97 Dinner Banquet

Banquet Talk:

How I Became Uncertain, Eugene Charniak

Sunday, August 3, 1997

Invited talk IV: Gaussian processes - a replacement for supervised neural networks?
David J.C. MacKay

Abstract: Feedforward neural networks such as multilayer perceptrons are popular tools for nonlinear regression and classification problems. From a Bayesian perspective, a choice of a neural network model can be viewed as defining a prior probability distribution over non-linear functions, and the neural network's learning process can be interpreted in terms of the posterior probability distribution over the unknown function. (Some learning algorithms search for the function with maximum posterior probability and other Monte Carlo methods draw samples from this posterior probability). In the limit of large but otherwise standard networks, Neal (1996) has shown that the prior distribution over non-linear functions implied by the Bayesian neural network falls in a class of probability distributions known as Gaussian processes. The hyperparameters of the neural network model determine the characteristic lengthscales of the Gaussian process. Neal's observation motivates the idea of discarding parameterized networks and working directly with Gaussian processes. Computations in which the parameters of the network are optimized are then replaced by simple matrix operations using the covariance matrix of the Gaussian process. In this talk I will review work on this idea by Neal, Williams, Rasmussen, Barber, Gibbs and MacKay, and will assess whether, for supervised regression and classification tasks, the feedforward network has been superceded.

Plenary Session V: Applications of Uncertain Reasoning

Break 10:40-11:00am

Panel Discussion: 11:00-12:00pm

Lunch 12:00-1:30pm

Plenary Session VI: Developments in Belief and Possibility

Break 3:00-3:30pm

Plenary Session VII: Topics on Inference

If you have questions or comments about the UAI '97 program, contact the UAI '97 Program Chairs: Dan Geiger and Prakash P. Shenoy. For questions about the UAI '97 conference, please contact the Conference Chair, Eric Horvitz.