BayesianNetworksLecture

Introduction

It has been argued that experts are not good at quantitative reasoning under uncertainty and therefore need support from software systems for this. Normative expert systems have been developed to leviate this problem.

Normative expert systems have the following properties over rule based systems:

instead of modelling the expert they model the domain;
they use classical probability calculus over noncoherent uncertainty calculus;
they are support systems for experts.

Bayesian Networks were introduced by Pearl (1986) as an alternative to expert systems and at that period of time the calculations for large networks were intractable. Since then Bayesian Networks have become viable tools for reasoning under uncertainty.

Bayesian Networks have been applied to identify blood types (Rasmussen 1995), interpretation of images (Binford et al. 1988, Levitt et al. 1990, Jensen et al. 1992), and for troubleshoting printing problems (Heckerman et al. 1995). For other examples see Pearl (1988) and Jensen (1996). As you can see Bayesian Networks have had a broad application.

The following are important issues in Bayesian networks:

CausalNetwork - A Bayesian Network, also called a Belief Network can represented causal relations and be known as a Causal Network.
BayesNetDseparation - An important concept to understand in the propagation of evidence is D-Separation.
BayesNetDefinition - The definition of a Bayesian network.
BayesNetComputationExample - An example of computing the joint probability of a graph.
BayesNetModeling - Modeling hints for Bayesian Networks.
UseOfBayesianNetworks - Uses for Bayesian Networks.
BayesNetAgents - Agents and Bayesian Networks

References

Binford, T., T. Levitt, W. Mann 1989. Bayesian inference in model-based machine vision. In Uncertainty in artificial intelligence, J. F. Lemmer & L. M. Kanal (eds). Amsterdam: Elsevier Science.

Heckerman, D., J. Breese, K. Rommelse 1995. Decision-theoretic troubleshooting. Communication of the ACM 38(3), 49-56.

Jensen, F. V., H. I. Christensen, J. Nielsen 1992, Bayesian methods for interpretation and control in multi-agent vision systems. In Proc. from applications of artificial intelligence X, K. W. Bowyer (ed.), 536-48. Orlando, FL: SPIE - The International Society for Optical Engineering.

Jensen, F. V. 1996. An introduction to Bayesian networks. New York, NY: Springer.

Levitt, T., T. Binford, G. Ettinger 1990. Utility-based control for computer vision, In Uncertainty in artificial intelligence, Schachter, Levitt, Kanal, Lemmer (eds), 407-22. Amsterdam:Elsevier Science.

Pearl, J. 1986. Fusion, propagation, and structuring in belief networks. artificial Intelligence 29(3), 241-88.

Pearl, J. 1988. Probabilistic reasoning in intelligent systems: networks of plausible inference, (Series in representation and reasoning). San Francisco, CA: Morgan Kaufmann.

Rasmussen, L. K. 1995. Bayesian network for blood typing and parentage verification of cattle. Dina research report no. 38. Aalborg University, Dept. of Mathematics and Computer Science.

Links to other Sources

Introduction and course notes [3][7]
Inductive Learning. [2][5][6]
Partially Observable Markov Decision Processes (POMDP)... for planning. [4]

		[AgentKnowledgeRepresentation | AgentKnowledgeRepresentation | CausalNetwork]

(last edited September 21, 2000)
Find Page by browsing or searching