MAI0096
Bayesian Networks/
Bayesianska nätverk

Number of credits: 6 hp

Examiner: John Noble

Course literature: Bayesian Networks; an Introduction by T. Koski and J. Noble (Wiley, 2009).

Course contents:

• Probabilistic reasoning: Bayes rule, Jeffrey's rule, Pearl's method of Virtual Evidence, Multinomial sampling and the Dirichlet integral.
• Conditional independence, graphs and d- separation, Bayesian networks, Markov equivalence for graph structures.
• Hard evidence, soft evidence, virtual evidence; Jeffrey's rule and Pearl's method.
• Decomposable graphs, junction trees and chain graphs.
• Learning the conditional probability potentials for a given graph structure.
• Learning the graph structure: the Chow-Liu tree, the Minimum Maximum Hill Climbing algorithm.
• Parametrising the network and sensitivity to parameter changes.
• Causality and Intervention Calculus.
• The junction tree and probability updating.
• Factor graphs and the sum product algorithm.

Organisation: 12 lectures, two lectures per week, in HT1.

Examination: 6 written / computer assignments, one per week, distributed during the course.

Prerequisites: Basic probability theory is necessary, some graph theory is helpful.

Page manager: karin.johansson@liu.se
Last updated: 2014-04-29