CMPE58K - Bayesian Statistics and Machine Learning

Fall 2011

Instructor: Ali Taylan Cemgil
(Volunteer) TA's: Arman Boyaci, Serhan Danis
Bogazici University,
Department of Computer Engineering,
Istanbul, Turkey

Announcements

  • (28.10) Assignment 3 is posted, See assignments tab,

  • (07.10) Second assignment is posted, See assignments tab,

  • (30.09) First assignment is posted fall2011-assignment1.pdf Due 5 Oct, 9:00

  • (30.09) There will be a PS on 5 Oct, 9:00-10:00 ETA A6

  • (26.09) The lectures will be in ETA A6 (Cmpe building), Wed 234 (10:00-13:00). Already 25 are registered so it won't be possible to change the time.

  • (26.09) Every week after the second, there will be a problem session before the lecture (9:00-10:00) where assignments will be discussed in detail.

Slides

  • Lecture00 Course Structure

  • Lecture01 Introduction, Probability tables,

  • Lecture02 Graphical Models, Independence, Conditional Independence, Additional slides from Bishop

  • Lecture03 Bayesian Learning, Applications

  • Lecture04 Exponential Family, Unvariate Probability Distributions, Conjugate Priors

  • Lecture05 Multivariate Distributions: Gaussian, Multivariate Bernoulli

  • Lecture06 Sequential Data, Markov models, Hidden Markov Models (HMM's), Inference, Forward Backward algorithm

  • Lecture07 Approximate inference, Variational Methods: Variational Bayes, ICM, Expectation Maximisation (EM), Variational Methods, AR Model Example

  • Lecture08 Correction Smoother, Junction tree algorithm

  • Lecture09 Linear Dynamical Systems (LDS's), Inference in LDS, Kalman Filter and Smoother

  • Supplement Changepoint models

Timetable

  1. Sep 28, Wed Introduction, Bayes Theorem, clustering, k-means, generative models

  2. Oct 05, Wed Derivation of the k-means algorithm from the generative model as ICM, Multinomial density, Indicator representations

  3. Oct 12, Wed Probability tables, Graphical models, Independence, Conditional Independence

  4. Oct 19, Wed Midterm I (ATC absent - WASPAA)

  5. Oct 26, Wed Hidden Markov Models, Filtering, smoothing, Forward pass, Viterbi path

  6. Nov 02, Wed

  7. Nov 09, Wed Kurban Bayrami

  8. Nov 16, Wed

  9. Nov 23, Wed

  10. Nov 30, Wed

  11. Dec 07, Wed

  12. Dec 14, Wed Midterm II (ATC absent - NIPS)

  13. Dec 21, Wed

  14. Dec 28, Wed

 
$\Large p(\lambda| x) = p(x|\lambda) p(\lambda) / \int d\lambda' p(x|\lambda') p(\lambda') $