Fundamentals of Bayesian Artificial Intelligence

Lectures
12 h

Recommended for
This course introduces Bayesian methods for artificial intelligence. No specific background is required; familiarity with AI concepts is presupposed.

The course was last given
New course.

Goals
To introduce the fundamental concepts of Bayesian reasoning, Bayesian networks for prediction and modeling, and the automated learning of Bayesian networks.

Prerequisites
Familiarity with artificial intelligence.

Organization
This course is organized as a series of lectures. Assessment will be via paper and programming exercises.

Contents
* Reasoning under uncertainty (RUU); Bayesian philosophy * Bayesian networks o History of RUU in AI o Properties of Bayesian nets o Evaluation methods: exact, stochastic and approximate o Dynamic Bayesian nets * Applications o Application tools o Medical decision making o Natural language generation (NAG) o Poker * Learning Bayesian networks o Probabilistic causal structure o TETRAD II o EM learning o Bayesian learning o MML learning (GAs, MCMC) * Evaluating machine learners o Orthodox evaluation o Bayesian confirmation theory o Kullback-Leibler distance o Information reward.

Literature
To be specified.

Teachers
Kevin Korb.

Examiner
Arne Jönsson.

Schedule
27-31 March 2000.

Examination
Exercises.

Credit
3 credits

Page responsible: Director of Graduate Studies