Unit 6: Uncertainty, Graphical Models, and Probabilistic Reasoning

Commentary

To solve problems in the real world, AI has to model uncertain information and knowledge, and be capable of performing reasoning, learning, and decision-making tasks under uncertainty. This unit introduces uncertainty, graphical models, and probabilistic reasoning, which have long been regarded as the foundation of AI principles and techniques. More importantly, these techniques have significantly contributed to recent AI advancements, and have gradually become a part of "mainstream AI". The concepts and principles presented in this unit also serve as background knowledge for studying the remaining units of this course.

Unit Purpose

When you complete this unit, you will be able to

• Define concepts about uncertainty, graphic models, and probabilistic reasoning.
• Explain the models and inference algorithms of graphical models, including Bayesian networks.
• Describe approximate inference in Bayesian networks, such as MCMC algorithms.
• Perform and implement the above probabilistic reasoning algorithms.
• Discuss other approaches to uncertain reasoning.

Section 1: Uncertainty, probability, and Bayesian networks
Section 2: Exact and approximate inference in Bayesian networks
Section 3: Other approaches to uncertain reasoning

Readings

Supplemental Unit Readings

Books:

Bishop, C. M. (2006). Pattern Recognition and Machine Learning. Springer. ISBN 0-387-31073-8. (Refer primarily to Graphical Models)

Robert, C. P., and Casella, G. (2004). Monte Carlo statistical methods (2nd ed.). New York, NY: Springer. ISBN 0-387-21239-6.

Pearl, J. (1988). Probabilistic Reasoning in Intelligent Systems: Networks of Plausible Inference. San Mateo, CA: Morgan Kaufmann.

Activities

• Explore other approaches to uncertain reasoning, and discuss them in the online course conference. Compare these approaches to Bayesian network reasoning.
• Explore MCMC research and applications in fields other than AI and computer science to see how general techniques can be applied to different areas.