Course Description:

This course aims to provide an introduction to the area of probabilistic models

based on graphs, and meanwhile to offer an overall view on examining the recent

trends of using graphical models in various application domains such as machine

learning, sensor network, bioinformatics, signal/multimedia processing, and

computer vision, etc. Through the study of key models, e.g., Bayesian networks,

state space models, Markov random fields, and conditional random fields, we will

investigate and test several popular algorithms related to exact or approximate

inference, including elimination algorithm, junction tree, sum-product and max-

product algorithms, loopy belief propagation, and variational methods. Students

taking the course are required to submit a final term project that emphasizes an

appropriate use of graphical model techniques on their choice of application.

Course Contents:

1. Basic probability theory and graph theory

2. Representations of graphical models

3. Learning from data

4. Methods for exact inference

5. Methods for approximate inference

6. Applications

Textbook:

Daphne Koller and Nir Friedman

"Probabilistic Graphical Models: Principles and Techniques"

The MIT Press, 1 edition, 2009

https://class.coursera.org/pgm-2012-002