- Handout 0 - about the course
- Lecture 1 class notes, 18 pages (postscript, pdf)
- Lecture 2 class notes, 11 pages, some color (postscript, pdf) -- Hidden Markov Models
- Lecture 2 class notes (extended), 19 pages, some color (postscript, pdf) -- more details on FB
- Lecture 3 class notes, 16 pages (pdf) -- Graphical Models of conditional independence
- Lecture 5 class notes, 7 pages (postscript, pdf) -- Variable elimination
- Lecture 6 class notes, 7 pages (postscript, pdf) -- Decomposable model, triangulation and the junction tree
- Lecture 7 class notes, 9 pages (postscript, pdf) -- The junction tree algorithm
- Factor Graphs and the Sum-Product Algorithm, 22 pages -- the sum-product algorithm for loopy graphs
- Lecture 9 and 10 class notes, 17 pages (pdf) -- Moments, enumeration, other operators
- "A Tutorial on Learning With Bayesian Networks" by David Heckerman, MSR-TR-95-06
- Lecture 11 pdf -- Learning the parameters of Bayesian networks
- A Wikipedia page with a graphical illustration of the Dirichlet prior (convex for N'_i>1 and concave for N'_i<1) is here
- A small matlab program that generates data from two independent binary variables X,Y and scores the models X Y and X->Y by various scores. It demonstrates the sensitivity of the "optimal" structure to the alpha (=N') parameter of the BDEu prior. You need to edit the alpha parameter inside the code to see this effect.
- "Inducing features of random fields" by Della Pietra, Della Pietra and Lafferty
- "Conditional random fields" by Lafferty, McCallum and Pereira
- "Discriminative training methods for hidden Markov models" by Collins
- Lecture 12, 16 pages (pdf) -- Tree graphical models: inference and ML estimation
"Learning with mixtures of trees" by Meila and Jordan, 2000
"An accelerated Chowand Liu algorithm: fitting tree distributions to high dimensional sparse data" by Meila, 1999
"Bayesian learning of tree distributions" by Meila and Jaakkola, 2006
|