IE598: Handouts

The lecture notes will be updated as the course proceeds. Here is a brief summary of the course

1. Overview (last updated 08.23.2016)

2. Graphical models (last updated 08.23.2016)

3. Markov property (last updated 08.31.2016)

4. Elimination algorithm/Belief propagation (last updated 09.15.2016)

5. Density evolution/LDPC code (last updated 09.15.2016)
   application: Crowdsourcing

6. Max-product algorithm (last updated 09.15.2016)

7. Gaussian graphical models (last updated 10.04.2016)

8. Restricted Boltzmann machines (last updated 10.11.2016)

9. Markov Chain Monte Carlo (last updated 10.20.2016)

10. Variational inference (last updated 11.03.2016)

11. Learning (last updated Fall 2012)

Final project

• project description from 2015 Spring

• Final Project for Fall 2016 is to read one of the following research papers and explain to the class in 25 minutes. You can use either the board or slides as you prefer. It will be nice to have a few pages of handout to share with the audience. The list of papers will be shared with the class on Piazza. Whoever claims a paper first in the reply gets it.

• Tuesday 1: Representational Power of RBM

• Tuesday 2: RBM for collaborative filtering

• Tuesday 3: Multi-prediction DBM

• Tuesday 4: Sum-product networks

• Thursday 1: Bethe partition function

• Thursday 2: Shallow vs. deep Sum-product networks

• Thursday 3: Sparse tensor graphical model

• Thursday 4: Robust Formant tracking

• Thursday 5: discrete graphical model

• Thursday 6: inference networks

• Thursday 7: expectation particle belief propagation

• Thursday 8: dynamic message passing

• Thursday 9: Wassestein training of RBM

• Thursday 10: motion style

• Thursday 11: CRF with Gaussian edge potential

• Thursday 12: Markov logic network

• Thursday 13: multimodal learning with DBM

• Friday 1: DBM

• Friday 2: CRF

• Friday 3: discriminative RBM

• Friday 4: modeling dcuments with DBM

• Friday 5: graphical lasso

• Friday 6: adaptive belief propagation

Overviews:

There are many notes that summarizes mostly intersecting and somewhat different areas in probabilistic graphical models. Depending on your background, some of the following references can prove to be extremely insightful.

• Daphne Koller, Nir Friedman, Lise Getoor and Ben Taskar 2007 Graphical models in a nutshell

• Michael I. Jordan 1997 An introduction to graphical models

• Michael I. Jordan, Yair Weiss Graphical models: probabilistic inference

Additional reading:

Markov properties

• Peter Bartlett 2003 ,Undirected Graphical Models: Chordal Graphs, Decomposable Graphs, Junction Trees, and Factorizations
  

• R. Shachter 1998, Bayes-ball: Rational pastime
  

Belief propagation

• C. C. Moallemi and B. Van Roy, 2007,Convergence of min-sum message-passing for convex optimization
  

• D. Gamarnik, D. Shah, and Y. Wei, 2011,Belief propagation for min-cost network flow: convergence and correctness
  

Gaussian graphical models

• Y. Weiss and W. T. Freeman, 2001, Correctness of belief propagation in Gaussian graphical models of arbitrary topology
  

• C. C. Moallemi and B. Van Roy, 2006, Consensus propagation
  

Variational inference

• J. S. Yedidia, W. T. Freeman, Y. Weiss, 2005, Constructing free-energy approximations and generalized belief propagation algorithms
  

• H. A. Bethe, 1935, Statistical Theory of Superlattices
  

• R. Kikuchi, 1951, A Theory of Cooperative Phenomena

Random processes on graphs

• A. Dembo, A. Montanari, 2010, Gibbs Measures and Phase Transitions on Sparse Random Graphs

Crowdsourcing

• Q. Liu, J. Peng, A. Ihler, 2012, Variational Inference for Crowdsourcing

• D. R. Karger, S. Oh, D. Shah, 2011, Budget-optimal Task Allocation for Reliable Crowdsourcing Systems