Introduction to Artificial Inteligence - Spring 2001

Reasoning under uncertainty

1. Acting under uncertainty
   • Handling uncertain knowledge: why?
   • Uncertainty and rational decisions - decision theory
   • Decision-theoretic agents
2. Review of probability theory
   • Prior probability
   • Conditional probability
   • Axioms of probability
   • Bayes rule
3. Representing knowledge in uncertain domains - Bayes networks
   • Probabilistic rules and graphical structure
   • Where to get the probabilities?
   • Bayes network - topology and distribution
4. Semantics of Bayes networks
   • Joint distribution of a Bayes network
   • Constructing networks from distribution
   • Compactness and node ordering
   • Alternate representation of conditional probability tables
   • Conditional independence in Bayes networks - d-separation
5. Computational problems of interest, and JavaBayes
   • Marginal probabaility computation - prior and posterior
   • Maximum probability explanation: MPE and MAP
   • Using JavaBayes to get the answers
6. Inference in Bayes networks
   • Probabilistic inference types: diagnostic, causal, intercausal, mixed.
   • Algorithm for answering queries in poly-trees (outline).
7. Inference in multiply-connected Bayes networks
   • Clustering schemes
   • Conditioning schemes
   • Symbolic schemes
   • Stochastic schemes
8. Other uncertainty formalisms
   • Dempster-shafer theory - representing ignorance
   • Fuzzy sets and logic - representing vaguness