- Days: Tuesday /
Thursday
- Time: 2:00 p.m. - 3:20
p.m.
- Room: CS 219
- Instructor: Rina Dechter
One of the main challenges in building intelligent systems
is the ability to reason under uncertainty, and one of the most successful
approaches for dealing with this challenge is based on the framework of
Bayesian belief networks. Intelligent systems based on Bayesian networks are
currently being used in a number of real-world applications including
diagnosis, sensor fusion, on-line help systems, credit assessment, and data
mining.
The objective of this class is to provide an in-depth exposition of
knowledge representation and reasoning under uncertainty using the framework
of belief networks. Both theoretical underpinnings and practical
considerations will be covered, with a special emphasis on constructing
graphical models and on exact and approximate inference algorithms.
Additional topics include learning belief network parameters from Data,
dynamic belief networks, reasoning about actions and planning under
uncertainty.
- Familiarity with basic
concepts of probability theory.
- Knowledge of basic
computer science, algorithms and programming principles.
- Previous exposure to
AI is desirable but not essential.
|
|
Topic
|
Date
|
|
Week 1
|
- Introduction:
Reasoning about beliefs using Logic and Probability (Pearl Chapters
1-2)
|
10/1
|
|
|
|
|
|
Week 2
|
- Bayesian network
representation I:Independence Properties Syntax and Semantics (Pearl Chapter
3)
|
10/8
|
|
|
- Bayesian network
representation II: Directed graphical models of independence
|
|
|
Week 3
|
- Bayesian network
representation III: Undirected graphical models of independence
|
10/15
|
|
|
- Knowledge Engineering
of Bayesian networks
|
|
|
Week 4
|
- Exact inference
using variable elimination methods
|
10/22
|
|
|
- Complexity of
inference tree-width
|
|
|
Week 5
|
- Distributed
inference I: Polytrees and jointrees (Pearl chapter 4)
|
10/29
|
|
|
- Distributed
inference II: Polytrees and jointrees
|
|
|
Week 6
|
- Conditioning
schemes Hybrids of inference and conditioning time-space tradeoffs
|
11/5
|
|
|
- Canonical models
& local representation techniques
|
|
|
Week 7
|
- Maximal a posteriori
computations (MPE, MAP) and their applications (Pearl chapter )5
|
11/12
|
|
|
- Approximate
inference: Stochastistic methods Variables elimination methods
Iterative belief propagation
|
|
|
Week 8
|
- Learning Bayesian
networks
|
11/19
|
|
Week 9
|
- Decision and
control I: Influence diagrams Maximizing expected utility Dynamic
Bayesian networks (Pearl chapter 6)
|
11/26
|
|
|
|
|
|
Week 10
|
- Assorted topics:
Markov decision processes Inference: Policy iteration, value iteration
Causality and action Students presentation
|
12/3
|
- Dechter, R., "Bucket
Elimination: A unifying framework for Reasoning."
- Judea Pearl, Probabilistic
Reasoning in Intelligent Systems. Heckerman & Breese, Causal
Independence for Probability Assessment and Inference Using Bayesian
Networks.
- Boutilier, Friedman,
Goldszmidt & Koller, Context-Specific
Independence in Bayesian Networks.
- Dechter, Bucket Elimination: A
Unifying Framework for Probabilistic Inference.
- Dechter, "AAAI98
tutorial on reasoning."
- Heckerman, A
Tutorial on Learning with Bayesian Networks.
- Kjaerulff, dHugin:
A Computational System for Dynamic Time-Sliced Bayesian Networks.
- Pearl, Causation,
Action and Counterfactuals.
- Dechter, Mini-buckets:
a general scheme for approximating inference.
- Darwiche, Recursive
Conditioning: Any-space conditioning algorithm with treewidth-bounded
complexity.
- Darwiche, Any-space
probabilistic inference.
- Darwiche, On the role of
partial differentiation in probabilistic inference.
- Horvitz, Breese,
Heckerman, Hovel & Rommelse. The Lumiere
Project: Bayesian User Modeling for Inferring the Goals and Needs of
Software Users.
- Binder, Murphy,
Russell. Space-efficient
inference in dynamic probabilistic networks.
- Russell, Binder,
Koller, Kanazawa. Local
learning in probabilistic networks with hidden variables.
- Dugad & Desai.
A Tutorial on
Hidden Markov Models.
- Friedman, Geiger,
Goldszmidt. Bayesian
Network Classifiers.
- Dechter, R., El
Fattah, Y., Topological
Parameters For Time-Space Tradeoff
- Gagliardi, F., Generalizing
Variable Elimination In Bayesian Networks
- Rish, I; Dechter, R,
AAAI
2000 Tutorial
- Judea Pearl. Probabilistic
Reasoning in Intelligent Systems. Morgan Kaufmann, 1990.
- Finn V. Jensen. An
introduction to Bayesian networks. UCL Press, 1996.
- Robert G. Cowell, A.
Philip Dawid, Steffen L. Lauritzen, David J. Spiegelhalter Probabilistic
Networks and Expert Systems Springer-Verlag, 1999
- Castillo, E.;
Gutierrez, J.M.; Hadi, A.S., Expert Systems and Probabilistic Network
Models, Springer-Verlag 1997
There will be homework assignments and students will also
be engaged in projects.
Homeworks and projects (50%),
midterm (50%)
|