Computational Inference

STAT 440 / 840, CM 461

News:

Winter 2006
Department of Statistics and Actuarial Science
University of Waterloo
 

Instructor: Ali Ghodsi
Room: RCH 207
Time: MWF 12:30-01:20
Office Hours:  Thursday  10am --11am or by appointment  (MC 6081G)

Description

This course will explore recent developments in the area of probabilistic / statistical modeling of complex, multivariate data. Until recently, the use of probabilistic and statistical models has been limited by the complexity of  exact  inference. However, recent advances from computing science have made many  inference tasks practical. This course will provide the fundamentals of graphical  models including Bayesian networks, Markov networks, hidden Markov models and  algorithms for inference with these models. Examples will be given of applications of these models in different areas.

Textbook

An Introduction to Probabilistic Graphical Models,
by Michael I.  Jordan

This textbook is not yet published but chapters from the preliminary draft will be distributed by the instructor.
 

Grading

Assignments:      60%
Project :             40%

Computing

Tentative  Outline

Assignments

Assignment 1

Assignment 2

Assignment 3

Project

Final project reports (up to 8 pages of PDF) are worth 40% of your final grade .You are encouraged to chose a topic related to your research area. However,  you cannot  borrow part of existing thesis work, nor can  re-use a project from another course.  

 The basic types of projects in this course are:   

• Develop a new algorithm. In this case, you will need to demonstrate (theoretically and/or empirically) why your technique  is better than other algorithms.
• Application of graphical models to some domain. This could either be your own research problem, or you could try reproducing results of someone else's paper. In this case, you may use an existing software package, such as  Bayes Net Toolbox for Matlab (BNT) ,  Netica, or other ones listed here.
• Implement an existing algorithm. In this case, you are strongly encouraged to implement the code in such a way as to make it  reusable by others.
• Application / survey of  a technique (or techniques) in computational inference . It includes theoretical investigation of a technique or problem  or  experimental investigation of a technique (or techniques) in an abstract setting.

 At the end, the project is  a chance to learn more about some sub- area of computational inference  that you might be most interested in, as well as create a potentially interesting system. However, you may benefit more from implementing an algorithm and doing some simulations rather than trying to read and summarize some state-of-the-art  papers.

To find an interesting topic  you might try visiting the library, surfing the web, or  even just thinking for a bit in order to identify an interesting task that you would like a computer to perform. I recommend start by reading the article entitled “Graphical Models” by  Michael Jordan. This article overviews the basic graphical model framework and basic inference algorithms, and provides several examples of real-life graphical models. As additional resources you will  find the following  journals and conferences quite useful:

Important Dates:

Lectures:

Latex template for Lecture notes