Probabilistic Graphical Models: Principles and Applications
Web page: ccc.inaoep.mx/~esucar/Clases-mgp/mgp.html
Dr. L. Enrique Sucar, Ciencias Computacionales, INAOE
1. Objective
Probabilistic graphical models
have become a powerful set of techniques used in several domains. This
course provides a general introduction to probabilistic graphical
models (PGMs) from an engineering perspective. It covers the
fundamentals of the main classes of PGMs: Bayesian classifiers, hidden
Markov models, Bayesian networks, dynamic and temporal Bayesian
networks, Markov random fields, influence diagrams, Markov
decision processes, and relational and causal PGMs; including
representation, inference, and learning principles for all the
techniques. Realistic applications for each type of model are also
covered.2. Outline
- Part I Fundamentals
- 1. Introduction
- 2. Probability theory
- 3. Graph theory
- Part II Probabilistic Models
- 4. Bayesian classifiers
- 5. Hidden Markov models
- 6. Markov random fields
- 7. Bayesian networks: representation and inference
- 8. Bayesian networks: learning
- 9. Dynamic and temporal Bayesian networks
- Part III Decision models
- 10. Decision graphs
- 11. Markov decision processes
Part IV Relational and causal models
- 12. Relational probabilistic graphical models
- 13. Graphical causal models
3. Bibliography
Text book:
- L.E. Sucar, Probabilistic Graphical Models: Principles and Applications, Springer, 2015.
Additional references:
- PGMs:
- D. Koller, N. Friedman, Probabilistic
Graphical Models, MIT Press, 2009.
- L.E. Sucar, E. Morales, J. Hoey, Decision Theory Models for Applications in Artificial Intelligence: Concepts and Solutions, IGI-Global, 2012.
- Probability:
- L. Wasserman, All of
Statistics, Springer,
2004.
- E.T. Jaynes, Probability
Theory: The Logic of Science,
Cambridge, 2003.
- BNs:
- J. Pearl, Probabilistic Reasoning in Intelligent Systems,
Morgan-Kaufmann,
1988.
- R. Neapolitan, Probabilistic Reasoning in Expert Systems,
Addison-Wesley,
1990.
- R. Neapolitan, Learning Bayesian Networks, Pearson,
2004.
- A. Darwiche, Modeling and
Reasoning with Bayesian Networks,
Cambridge, 2009.
- Statistical prespective:
- C. Borgelt, R. Kruse, Graphical Models, Wiley, 2002.
- Whittaker, Graphical Models, Wiley, 1990.
- Differenet approaches on uncertainty:
- Shafer y Pearl (Eds.), Readings in Uncertain Reasoning,
Morgan-Kaufmann,
1990.
- Decision models:
- F. Jensen, Bayesian Networks and Decision Graphs,
Springer,
2001.
- R. Borrás, Análisis de incertidumbre y riesgo para la toma de decisiones, Orizaba, Mexico:
Comunidad Morelos, 2001.
- M Putterman, Markov Decision Processes, Wiley, 1993.
- Classifiers, learning:
- D. Michie, D.J. Spiegelhalter, C.C. Taylor, Machine
Learning,
Neural
and Statistical Classification, Elis Horwood, 1994.
- B. Sierra Araujo (Ed.), Aprendizaje automático:
conceptos
básicos y avanzados, Pearson, 2006.
- Causality:
- J. Pearl, D. Mackenzie, The Book of Why. Basic Books, 2018.
- J. Pearl, Causality: models, reasoning and inference. Cambridge, 2009.
- P. Sprites, C. Glaymour, R Scheines, Causation, Prediction and Search. MIT Press, 2001.
- Applications:
- O. Pourret, P. Naim, B. Marcot (Eds.), Bayesian Networks: A
Practical Guide to Applications, Wiley, 2008
4. Main journals and conferences
- International Journal of Approximate Reasoning
- International Joint Conference on Artificial Intelligence (IJCAI)
- Uncertainty in Artificial Intelligence (UAI)
- Probabilistic Graphical Models (PGM)
- FLAIRS - Uncertain Reasoning Track
- Information Processing and Management of Uncertainty (IPMU)
- Symbolic and Quantitative Approaches to Uncertainty (ECSQAU)
- Neural Information Processing (NIPS)
5. Course policies
5.1 Prerrequisites
There are no particular requisites for the course. Basic knowledge of math and computing is assumed.
5.2 Course notes
All class notes will be available in the course web page.
5.3 Homework, projects and exams
Homeworks:
will be assigned each week, which in general are not required to be
hand in. I will ask some of the students to solve the excercises
in class. In a few cases I will tell you if you have to deliver a
report. Homeworks are individual.
Final Project:
by the midde of the course a project will be assigned (could be
individual or in two persons teams). The project should implement /
apply a PGM technique to certain problem. It consists of 3 phases:
- Proposal: short (1 page) description of the project objective, methodology and expected results.
- Preliminary: a
description of the work in the format of a scientific paper including
introduction, related work, methodology, results and conclusions (max
10 pages in Springer LN format).
- Final report: final paper taking into accound the feedback.
Exam:
there will a mid term exam which will include the topics cover until
the previous class. It consists of two parts: theory and problems.
5.4 Evaluation
- Homework - 20%
- Exam - 40 %
- Final project - 40 %
5.5 Individual advise
By appointment
5.6 Professor
Dr. L. Enrique Sucar S. (office / extension 8208)
e-mail: esucar (AT) inaoep.mx
web page: http://ccc.inaoep.mx/~esucar