Syllabus
Probabilistic Models of
Human and Artificial Intelligence

CSCI 7222 / CSCI 4202
Spring 2007

Tu, Th 14:00-15:15
ECCR 108

Instructor

Professor Michael Mozer (mozer@cs.colorado.edu)
Department of Computer Science
Engineering Center Office Tower 7-41
(303) 492-4103
Office Hours:  Tu, Th 12:30-13:30

Course Objectives

The goal of the course is to understand the latest advances in probabilistic models in artificial intelligence and cognitive science.  The course participants are likely to be a diverse group of students and faculty, some with primarily an engineering/CS focus and others primarily interested in cognitive modeling (building computer simulation and mathematical models to explain human perception, thought, and learning).

Probabilistic models have established a firm foothold in AI and ML over the past decade, but only in the past few years have cognitive scientists realized their potential for explaining human cognition.  One neat thing about probabilistic models is that they straddle the gap between AI/ML and cognitive science:  The same methodology is useful for both understanding the brain and building intelligent computer systems.  Indeed, for many of the papers we'll discuss, the models contribute both to machine learning and to cognitive science.  Whether your primary interest is in engineering applications of machine learning or in cognitive modeling, you'll see that there's a lot of itnerplay between the two fields.

Prerequisites

The course is open to any students who have some background in cognitive science or artificial intelligence.  Some background in proababilty and statistics will be helpful, but iis not essential as long as you are willing to learn.

Course Requirements

Readings

In the style of graduate seminars, your primary responsibility for the course will be to read the series of papers before class and be prepared to come into class to discuss the paper (asking clarification questions, working through the math in the paper, relating the paper to other readings, critiquing the paper, presenting original ideas related to the paper).

Written Commentaries

For some of the readings, I'll ask you to write a one-page commentary on the paper, The commentary consists of approximately one page of comments, questions, or critiques of the assigned reading(s) for that class. This page will be due the day of class, and can include one or more of the following:
These commentaries are intended to promote careful thought about a paper before the session in which it is discussed. The point is not to give you more busy work, but rather to encourage you to jot down notes and questions as you read the papers. They will not be accepted after the class in which the paper is discussed.

I won't ask you to write commentaries on readings that are primarily tutorial/review.

Presentation

You are required to present  a share of the papers during the course of the semester.  The presentation is meant to be a summary of the paper and its main ideas.  Ideally, two class members will collaborate to do each presentation, allowing you to work through the papers together.  I expect grad students to do twice the presentations that undergrads do.  My guess is that grad students will do about 3 presentations and undergrads will do 1, but that depends on class size.

Research Project

All students enrolled in the course will implement and test a probabilistic model and will write up a summary of their research for the end of the semester.  For grad students, ideally the work you'll do will be useful in your own research.  For undergraduates, you should start by implementing a model that we will discuss during the semester, replicating the results in the paper, and then going on to develop your own application of the model.

Semester Grades

Grades will be based roughly on the following: project 20%, oral presentation 20%, class discussions 10%, written commentary on papers 50%.

Class-By-Class Plan and Course Readings

Classes marked by (*) do not require commentaries.

Date Activity Reading Lecture notes presenter
Jan 16 introduction (*) Mozer, Jones, & Shettel (2007) download Mozer
Jan 18 introduction (*) Chater, Tenenbaum, & Yuille (2006)
Tenenbaum (1999)
download Mozer
Jan 23 a primer on probabilistic inference (*) Griffiths & Yuille (2006) download Mozer
Jan 25 Bayesian Ockam's razor (*) Jefferys & Berger (1991) download Mozer
Jan 30 inference in graphical models (*) Jordan & Weiss (2002)
download Mozer
Feb 1 MCMC (*) Andrieu et al. (2003) download Mozer
Feb 6 learning in Bayes nets (*) Heckerman (1995), pp. 1-24 download Mozer
Feb 8 learning in Bayes nets (*) Heckerman (1995), pp. 25-57 download Mozer
Feb 13 topic model Griffiths & Steyvers (2002) download Mozer
Feb 15 attention Mozer, Shettel, & Vecera (2006) download Mozer
Feb 20 latent dirichlet allocation Blei, Ng, & Jordan (2002)
longer version available as Blei, Ng, & Jordan (2003)
download Mozer
Feb 22 hierarchical topic model Blei, Griffiths, Jordan, & Tenenbaum (2004) Hadjar Homaie
Feb 27 integrating topics and syntax Griffiths, Steyvers, Blei, & Tenenbaum 2005

Noralie Sarver
Mar 1 NO CLASS


Mar 6 topic and role discovery in social networks McCallum, Corrado-Emmanuel, & Wang 2005 download Mozer
Mar 8 motion illusions as optimal percepts Weiss, Simoncelli, & Adelson (2002) Ted Fisher
Mar 13 vision as Bayesian inference Yuille & Kersten (2006) Kristopher Nuttycombe
Mar 15 Gaussian Processes (*) Williams (1997) download Mozer
Mar 20 Gaussian Processes (Computing with Infinite Networks) Williams (1997b)
Sam Reid
Mar 22 Probabilistic computation in spiking populations Zemel, Huys, Natarajan, & Dayan (2005) Jason
Boorn
Mar 27, 29 SPRING BREAK
Apr 3 Hidden Markov models (*) Gharamani (2001) Michael Otte, Scott Richardson 
Apr 5 Bayesian online changepoint detection Adams & MacKay (unpublished) Brian Loughry
Apr 10 Prediction and change detection Steyvers & Brown (2006) Michele Samorani
Apr 12 Hierarchical Dirichlet processes (and hierarchical beta processes) Teh, Jordan, Beal, & Blei (2003) Abhishek Jaiantilal
Apr 17 Theory-based causal inference Tenenbaum & Griffiths (2003) Chris DiHoust
Apr 19 Learning domain structures Kemp, Perfors, & Tenenbaum (2004) Mark Lewis-Prazen
Apr 24 Context sensitive induction Shafto et al. (2005) Kai Ching
Apr 26 Infinite relational models Kemp, Tenenbaum, Griffiths, Yamada, Ueda (2006) Richard Bell
May 1 Combining causal and similarity based reasoning Kemp, Shafto, Berke, & Tenenbaum (2007) Andrew Boehm
May 3 Sensorimotor control Koerding, Tenenbaum, & Shadmehr (2007)
Ben Pearre, Matt Wilder
May 5,  10:30 a.m. - 1 p.m. FINAL EXAM SLOT Not planning to hold class during exam period

Queue

The queue is a list of papers that we haven't yet scheduled but that I hope to cover during the semester

LANGUAGE
Conditional random fields
probabilistic models of language processing and acquisition  Chater & Manning (2006)
PERCEPTION  
visual adaptation as optimal information transmission Wainwright
MOTOR CONTROL
Bayesian decision theory in sensorimotor control Koerding & Wolpert (2006)
CONCEPT LEARNING
hierarchical topic model -- Griffiths Steyvers 2004
Probabilistic inference in human semantic memory Steyvers, Griffiths, & Dennis (2006)
REASONING/INFERENCE
Theory-based Bayesian models of inductive learning and reasoning Tenenbaum, Griffiths, & Kemp (2006)
CATEGORY LEARNING
Discovering multiple structures that capture diffrent subsets of features   Shafto et al.
Structure learning Kemp & Tenenbaum
OTHER
everyday prediction problems (Griffiths & Tenenbaum, 2006)
Neural Models of Bayesian Belief Propagation (Rao, 2006)
hyperhyperhyperparameters -- how far back can we push Bayesian inference

Interesting Links

Tutorials


Josh Tenenbaum's Bayesian models tutorial at NIPS 2006
Carl Rasmussen's Gaussian Processes tutorial at NIPS 2006
Jordan tutorial on hierarchical Dirichlet processes

Modeling tools

Topic modeling toolbox
UCLA's samiam
Murphy's Bayes net toolbox
BUGS
OpenBayes
Orange
Chris DeHoust comments on software