Potential report topics, Put your name by the one you want. Attach your report in Wiki form under the indicated link. Note that your report should be suitable for inclusion in Wikipedia. If one already exists you have to option to make a significant addition to the report that is there. You are permitted to actually put your report in Wikipedia and make the world a better place, but you might want to have the rest of the class critique it before you do so.

You are also required to make an oral report. Your reports must be understandable to your peers in the class. Since you will have little time for your presentation, please make a solid, coherent starting point rather than trying to tell us a little about all of the subject. The written report will be more complete.

NOTE ALSO that written reports are due a week before the last day of class.

• Undirected Models (Chapter 4) Undirected Models - Benjamin Sellers

• Variable Elimination (Chapters 9) Variable Elimination - Derrall Heath

• Clique Trees (Chapters 10) Clique Trees - Yanling Yang

• Slice Sampling Slice Sampling - Rob Smith

• Other inference techniques (Chapter 11 except for Variational)

• Variational Bayes (Parts of Chapter 11 and elsewhere) Variational Bayes - Kelv Cutler

• MAP Inference (Chapter 13) MAP Inference - Richard Morris

• Expectation Maximization (EM) (Section 19.2) EM - Matt Gardner

• Structure Learning (Chapter 18 and 19.4) Structure Learning - Mike Roscheck

• Structure Scores (Chapter 18) Structure Scoring - Paul Bodily

• Causality (Chapter 21) Causality - Aaron Dennis

• EVSI for the Natural Exponential Family (Chapter 8, also see Introduction to statistical decision theory John W. Pratt, Howard Raiffa, Robert Schlaifer. Cambridge, Mass. MIT Press, 1995.) General EVSI Decision-Theoretic Graphical Models - Josh Hansen

• Chinese Restaurant Process Chinese Restaurant Process - Yisong Guo

• Some other inference or learning procedure (check with me first, then add a link)

• An application of some concept from class (check with me first, then add a link)