21. Introduction to Bayesian Learning
Monday, Tutorial, 3:45 - 5:30 pm
Room 515A
Level: Intermediate

Sophisticated computer graphics applications require complex models of appearance, motion, natural phenomena, and even artistic style. Such models are often difficult or impossible to design by hand. Recent research demonstrates that, instead, we can "learn" a dynamical and/or appearance model from captured data, and then synthesize realistic new data from the model. For example, we can capture the motions of a human actor and then generate new motions as they might be performed by that actor. Bayesian reasoning is a fundamental tool of machine learning and statistics, and it provides powerful tools for solving otherwise-difficult problems of learning about the world from data. Beginning from first principles, this course develops the general methodologies for designing learning algorithms and describes their application to several problems in graphics.

Prerequisites
Familiarity with linear algebra, calculus, and computer graphics.

Intended Audience
Computer graphics researchers and practitioners working on data-driven computer graphics problems, such as animating shape and motion from video or animating from motion capture.

Organizer and Lecturer
Aaron Hertzmann
University of Toronto

Schedule

3:45	Introduction The Future of Graphics: Data-driven Analysis and Synthesis The Need for Bayesian Reasoning Hertzmann
4	Fundamentals of Bayesian Probabilistic Reasoning: Classical (Aristotelian) Logic and its Limitations Cox Axioms Bayes Rule Prediction and Parameter Estimation Learning Multinomials and Gaussians Relation to Least-squares Fitting and Frequentist Methods Hertzmann
4:30	How to Design Learning Algorithms: Generative Models The Overfitting Problem in MAP Learning and Least-squares Marginalization, The EM Algorithm at a Glance Example: Probabilistic PCA Example: Automatic Non-rigid Modeling From Video Hertzmann
5:15	The Summary and Conclusions: Pros and Cons of the Bayesian Approach Audience Questions Hertzmann