Probabilistic Graphical Models  10-708, Fall 2005 School of Computer Science, Carnegie-MellonUniversity

 

Module	Material covered	Online material and links	Dates and Instructor
Module 0: introduction (1 Lectures)		9/12: Introduction 9/12: The Basics 9/12: The Basics - Annotated slides	Sep 12:  Eric Xing & Carlos Guestrin
Module 1: Fundamentals of Bayesian Networks: representations, semantics and inference (7 Lectures)	Representation Semantics Exact inference Variable elimination  Junction trees	9/14 BN Semantics  9/14: BN Semantics - Annotated slides 9/14: JavaBayes - BN applet 9/19: BN Semantics2 (now in pdf) 9/19: BN Semantics2 - Annotated slides 9/21: BN Semantics3, Inference, Complexity (pdf) 9/21: BN Semantics3, Inference, Complexity - Annotated slides 9/26: Variable elimination 9/26: Variable elimination - Annotated slides 9/28: Variable elimination part2, Clique trees 9/28: Variable elimination part2, Clique trees - Annotated slides 10/03: Clique trees part2 10/03: Clique trees part2 - Annotated slides 10/05: Context-specific independence, parameter learning: MLE  10/05: Context-specific independence, parameter learning: MLE - Annotated slides	Sep 14: Carlos Guestrin Sep 19: Carlos Guestrin Sep 21: Carlos Guestrin Sep 26: Carlos Guestrin Sep 28: Carlos Guestrin Oct 3: Carlos Guestrin Oct 5: Carlos Guestrin
Module 2: Learning directed and Undirected Graphical Models (5 Lectures)	Model formalisms Exponential familiy models Generalized linear models Hierarchical Bayesian models Undirected graphical models and Markov random fields MLE, KL, and Max-Entropy principle Mixtures and conditional mixtures Learning completely observed graphical models Maximal likelihood and Bayesian parameter extimation Gradient algorithms IRLS algorithm IPF for tabular MRFs GIS for general MRFs Structural Learning Learning partially observed graphical models EM for MLE parameter estimation	10/10: MLE and Bayesian Learning of Simple BNs 10/12: Learning Exponential Family models, Generalized Linear Models and and general BNs 10/17: Undirected Graphical Models 10/19: Learning Completely Observed Undirected Graphical Models 10/24: Structural learning of completely observed BNs 10/24: Structural learning of completely observed BNs - Annotated slides 10/26: Structural learning of completely observed BNs 2 - Annotated slides 10/26: Learning partially observed BNs	Oct 10: Eric Xing  Oct 12: Eric Xing Oct 17: Eric Xing Oct 19: Eric Xing Oct 24: Carlos Guestrin Oct 26: Eric Xing
Module 3: More complex GMs: canonical temporal models, latent space models, hybrid models (3 Lectures)	Hidden Markov model Forward-backward Baum-Welch  Factor analysis Matrix algebra for multivariate Gaussian EM State space models Kalman filter RTS smoother Factorial HMM and Swithcing SSM Latent mixture membership models	10/31: Hidden Markov Models 11/02: Factor Analysis and State Space Models 11/07: Complex Models and Approximate Inference	Oct 31: Eric Xing Nov 2: Eric Xing Nov 7: Eric Xing
Module 4: Approximate Inference  (2 Lectures)	Sampling,  MCMC: MH, Gibbs Variational inference Mean field, Kikuchi, GBP	11/09: MCMC  and Belief Propagation  11/14: Variational Inference	Nov 9: Eric Xing Nov 14: Eric Xing
Module 5: Advanced Topics  (5 Lectures)	Dynamic models Kalman filter Linearization Switching Kalman filter Assumed density filtering DBNs BK RPMs (tentative) Applications  Information retrieval Bioinformatics Signal processing  Causality Decision-making	11/16: Dynamic models 1 - Linearization, Switching Kalman filter, Assumed density filter 11/16: Extra reference: Uri Lerner's thesis 11/16: Dynamic models 1 - Linearization, Switching Kalman filter, Assumed density filter - Annotated slides 11/21: Dynamic models 2 - Assumed density filter, DBNs, BK 11/21: Extra reference: Boyen and Koller 1998, 1999 11/21: Extra reference: Paskin 2003 11/21: Dynamic models 2 - Assumed density filter, DBNs, BK - Annotate slides 11/28: Models with Higher-level Structures 11/30: Extra reference: An Introduction to Causal Inference, (1997), R. Scheines 11/30: Extra reference: Causal Inference (2004). Spirtes, P., Scheines, R.,Glymour, C., Richardson, T., and Meek, C. 11/30: Prof. Schienes' Causality slides 11/30: Prof. Schienes' cute stained teeth example of intervention in causality slides - Flash required 11/30: Prof. Schienes' Tretad - the causality software he showed in class	Nov 16: Guestrin Carlos Nov 21: Guestrin Carlos Nov 28: Eric Xing Nov 30: Guest lecturer: Richard Scheines on causality!  Dec 2: Poster session Dec 5: Carlos Guestrin Dec 7: NO CLASS
Final Exam	All material thus far