This syllabus is tentative and subject to updates.
Bayesian Statistics: Calendar/Syllabus/Plan
ISyE8843: Bayesian Statistics
Brani Vidakovic, Groseclose 320
Office Hours : 12:00PM - 13:30PM Monday/Wednesday.
Class meets: 10:05 am - 10:55 am, Groseclose 304.
Textbook: Various texts in the field.
Prerequisites: No formally imposed prerequisites. Students are
expected to have exposure to classical inferential statistics (at least
ISyE2028 level) and a solid quantitative background (math and programming).
Weekly quiz on Fridays, (the lowest score will be dropped).
No makeups without a GOOD reason.
Exams include an in-class Midterm
(Wednesday, 10/15/2004 in class) and a Final (Monday, 12/6/2004; 2:50 - 5:40).
The exams will be open-book, open-notes.
In addition students will work on a project. Deliverable is an electronic copy
of the project; also students will make presentation on their projects during last
two class periods.
Quizzes 20%, Midterm 20%, Class Project and its Presentation 30%, Final 30%.
Academic Honor Code: Please
familiarize yourself with the
Georgia Tech Honor Code.
Week 1, 8/16-18-20
Review of Probabilities, Conditional Probabilities and
- Nature of Bayesian Inference. History. Appetizers.
Computation and Software.
- Philosophy (Berger and Wolpert: Likelihood Principle)
Week 2, 8/23-25-27
Single Parameter Models (3)
- Priors. Bayes Rules. Noninformative Priors.
Week 3, 8/30; 10/1-3
Bayesian Models (3)
- Bayesian Inference. Credible Sets, Testing Hypotheses, Bayes Factors.
- Utility/Loss. Bayesian Decision Theory. Frequentist -- Bayes Agreement.
Week 4, 9/8-10 (10/6 School Holiday)
- Families of Priors. Gamma-minimaxity.
Week 5, 9/13-15-17
Empirical Bayes Methods. ML II. Hierarchical Models. Hidden Mixtures.
Week 6, 9/20-22-24
Model Checking and Improvement. Modeling Accounting for Data
Week 7, 9/27-29; 10/1
- Pre-MCMC techniques.
- Random Number Generation. MC techniques.
Week 8, 10/4-6-8
MCMC Methods. General Theory.
- Markov Chains. Metropolis-Hastings Algorithm. Different Flavors of
MH Algorithm. Gibbs Sampler.
Week 9, 10/11-13-15 (10/15 Midterm)
Basics of WinBugs.
- Interplay of Matlab and WinBugs.
Week 10, 10/20-22 (10/18 Fall Break)
Various MCMC Topics.
- Rao-Blackwellization. Reversible Jump MCMC. Convergence Disgnostics.
Metropolis within Gibbs.
Week 11, 10/25-27-29
Missing Data Models.
- Hidden Markov Models. Finite Mixtures of Distributions.
Change Point Analysis.
Week 12, 11/1-5
Advanced Modeling in Various Scientific Fields and Engineering. More BUGS.
- Biology, Genetics, Medicine, Finance, Internet (Bayes Spam)
Week 13, 11/8-10-12
Bayesian Methods in Signal Processing and Image Analysis.
- Particle Filters. Pixel Level Models (Ising, Potts), Wavelets.
PET & SPECT. Template Models.
Week 13, 11/8-10-12
Linear and Non-linear Regression. Bayesian Networks. (3)
- Engineering Applications. Hierarchical Modeling in Regression.
Conditional Models. Causality. Wavelets
Week 14, 11/15-17-19
Bayesian GLM's. Categorical Data. (3)
- Categorical Data. Jim Albert's Matlab Suite and Book of Val Johnson
and Jim Albert, Ordinal Data Modeling .
Week 15, 11/22-24 (11/26 School Holiday)
Bayesian Model and Variable Selection (2)
- Stochastic Search. Mixing and Averaging Models. Wavelet Shrinkage Application.
Week 16, 11/29; 12/1-3
Bayes Discrimination (1), Student Projects (2)
- Classification. Combining Classifiers. Data Mining.
Last Updated: Sunday, July 11, 2004.
Back to the main course page.