Monash University > School of Computer Science and Software Engineering > CSE458> Lectures

CSE458 Bayesian Models

**TENTATIVE TIMETABLE**
Week	Week Commencing	Lectures Fridays 11.00 - 1.00 Lecture Theatre S9	Assessment (Follow link for further information regarding submission times etc.)
1	28/2/05	Lecture 1: Introduction Lecture 2: More Decision Analysis
2	07/03/05	Lecture 3: Utility Lecture 4: Probability
3	14/03/05	Lecture 5: Bayesian Networks	14/03/05	Submit Exercise 1
4	21/03/05	Good Friday (Holiday)
5	28/04/05	Mid Semester Break
6	04/04/05	No lecture (Professor Webb away from University)	04/04/05	Submit Exercise 2 Note re exercise 2: please treat the question as if only one of the two experiments can be performed, seismic soundings or the experimental device
7	11/04/05	No lecture (Professor Webb away from University)
8	18/04/05	Lecture 6: Decision Networks Lecture 7: Evaluation of Decision Networks	22/04/05	Submit Exercise 3
9	25/04/05	Lecture 8: Dynamic Networks Lecture 9: Independence
10	02/05/05	Lecture 10: Inference in Polytree Networks Lecture 11: Inference in Multiply Connected Networks
11	09/05/05	Lecture 12: Knowledge Engineering	13/05/05	Submit Exercise 4
12	16/05/05	Lecture 13: TBC (Allowance for catch-up if required)	20/05/05	Submit Assignment
13	23/05/05	No lecture
14	30/05/05	No lecture (Professor Webb away from University)

Details of Lectures

Introduction

Overview of the course, introduction to decision analysis, decision flow diagrams, Bayes rule, averaging out and folding back.

(ps)

(pdf)

More Decision Analysis

Uncertain payoffs, sampling costs, value of information, biased measurements, extensive form of decision analysis, strategies, and normal form of decision analysis.

(ps)

(pdf)

Utility

Lotteries, utilities, utility curves, maximization of expected utility, risk aversion, risk control, and buying and selling prices.

(ps)

(pdf)

Probability

Interpretation of probability, axioms of probability, independence, conditional probability, applications of Bayes' Rule, random variables, joint probabilities, and marginal distributions.

(ps)

(pdf)

Bayesian Networks

Networks, semantics of Bayesian Networks, representation of joint probabilities, compactness, node ordering, and inference.

(ps)

Decision Networks

Definition of a Decision Network, Semantics, and Decision Theory using Decision Networks.

(ps)

(pdf)

Alarm Network (Bayesian Network)

Alarm Network (Correct Decsion Network)

Alarm Network (Incorrect Decision Network)

Urn Network for e0

Urn Network for e1

Urn Network for e2

Incorrect Urn Network for es

Urn Network for es

Total Urn Network (version 1)

Total Urn Network (version 2)

Total Urn Network (version 3)

Evaluation of Evaluating Decision Networks

Overview of different approaches to evaluate Decision Networks.

(ps)

(pdf)

Dynamic Networks

Dynamic Bayesian Networks, Hidden Markov Models, Filtering, Prediction, Smoothing, most likely explanation, Dynamic Decision Networks, Markov Decision Processes.

(ps)

(pdf)

Independence

Semantics of Bayesian Networks, Construction of Bayesian Networks, Makov Blanket, Blocking, Independence, and Inference in Dynamic Bayesian Networks.

(ps)

(pdf)

Inference in Polytree Networks

Backward-chaining and Message passing

(ps)

(pdf)

(ppt)

Backward-chaining algorithm

(ps)

Inference in Multiply Connected Networks

Overview of different approaches to inference in Multiply Connected Networks

(ps)

Papers and other material of interest

Message Passing: Richard E. Neapolitan (2003), Learning Bayesian Networks, Chapter 3.

Clique-tree Inference: Inference in Belief Networks: A Procedural Guide, by Cecil Huang. International Journal of Approximate Reasoning 15:225-263, 1996.

Likelihood weighting: An Optimal Approximation Algorithm for Bayesian Networks by Paul Dagum and Michael Luby.

MCMC: Introduction to Monte Carlo Methods by David MacKay.

Building Networks

Topics:Knowledge acquisition, constructing models, simplifying local models, discrete distributions, continuous distributions, model evaluation, and case studies.

(ps)

(pdf)

Last modified 9/2/2005