Stanford University EE 374
Inference in Graphical Models
Winter 2006–2007

Course Information

Class Times and Locations

Course Description

Graphical models are a unifying framework for describing the statistical relationships between large sets of variables. The course addresses the basic question of computing the marginal distribution of one, or a few, of such variables. The focus is on sparse graph structures and theoretical analysis. Topics include: message passing algorithms; belief propagation; survey propagation; correlation decay; density evolution; distributional recursions; the cavity method; sparse graph codes; multi-user detection; random combinatorial optimization (random K-satisfiability).

Prerequisites

EE 278 or STAT 116 or CS 228 (required), EE 376A or STAT 217/218 (recommended)

Instructors

Grading

There will be several homeworks and a final project. In each class a volunteer will be solicited to scribe the lecture in Latex. In grading weight will be assigned according to: 35% for homework; 35% for project; 30% for notes and class participation.

Handouts

Handout Posted
Syllabus [pdf] 1/11
Project suggestions [pdf] 1/18
BP for pairwise graphical models [pdf] 1/18

Lecture Notes

Each student is required to scribe lecture notes for at least one class. Volunteers will be solicited at the beginning of each lecture. Students may work in pairs as scribes. It is suggested to use this style file in latex for the notes. Email the completed notes in PDF format along with the source files (since they may be useful to the next scribe) to ciamac@stanford.edu. The notes are due before the following lecture.

Lecture Posted Scribe Prof. Montanari's Notes
Lecture 1 [pdf] [source] 1/15 Ciamac Moallemi  
Lecture 2 [pdf] [source] 1/21 Esteban Arcaute  
Lecture 3 [pdf] [source] 1/22 Fernando Amat and Argyris Zymnis  
Lecture 4 [pdf] [source] 1/24 John Duchi  
Lecture 5 [pdf] [source] 1/30 Krishnamurthy Iyer  
Lecture 6 [pdf] [source] 2/5 Vahideh H. Manshadi  
Lecture 7 [pdf] [source] 2/13 Sewoong Oh  
Lecture 8 [pdf] [source] 2/13 Yi Lu  
Lecture 9 [pdf] [source] 2/19 Arash Asadpour and Dominic DiPalantino [pdf]
Lecture 10 [pdf] [source] 2/18 Joëlle Skaf [pdf]
Lecture 11 [pdf] [source] 2/19 Erick Delage [pdf]
Lecture 12     [pdf]
Lecture 13 [pdf] [source] 3/7 Farshid Moussavi [pdf]
Lecture 14 [pdf] [source] 3/5 Sewoong Oh [pdf]
Lecture 15 [pdf] [source] 3/19 Farshid Moussavi [pdf]
Lecture 16 [pdf] [source] 3/21 Dominic DiPalantino [pdf]
Lecture 17 [pdf] [source] 3/24 Arash Asadpour [pdf]
Lecture 18     [pdf]

Homeworks

Students are encouraged to work on homework problems in groups but must write up their own solutions. When writing up solutions, students should write the names of people with whom they discussed the assignment.

Homework Posted Due
Homework 1 [pdf] 1/18 2/1
Homework 1 Solutions [pdf] 2/19
Homework 2 [pdf] 2/22 3/8

Course Readings

The course readings will consist of a series of papers that will be distributed throughout the term. Reading the papers is not required but being curious and browsing through them is strongly recommended. The project will require studying a few of them. Access to these papers is restricted to the Stanford community, and requires authentication from outside the stanford.edu domain.