Probabilistic Graphical Models

Google groups link: CS6730 : Probabilistic Graphical Models

Office hours: Tuesdays 2 to 4pm, Fridays 10 to 12AM.

Class hours:

Tuesday - 4:50 to 5:40 PM

Wednesday - 11 to 11:50 AM

Thursday - 9 to 9:50 AM

Friday - 8 to 8:50 AM

Homeworks:

1. Submit tex source files, pdfs and Python code on Moodle. You may use ipe or similar tools for generating pdf figures if required.
2. There will be small amount of programming required here for some of the questions, maybe about 30 lines of Python code per homework.
3. I expect academic honesty from the students and to refrain from copying.

Project/Programming assignments:

1. This will involve implementing many of the inference/sampling/learning algorithms learnt in class, and testing on synthetic models and data.
2. The details of the project will be revealed as soon as the content required for implementing it is taught in class.

Evaluation:

Homeworks: 20

Project: 20

Quiz 1: 15

Quiz 2: 15

Final : 40

Class participation: 10

Total= max(H,P)+Q1+Q2+F+C

Announcements:

25/01 : Quiz 1 will be scheduled on Feb 20, Tuesday.

25/01: HW1 is up. Due Feb 7th. (pdf) (tex) (template code) (solution pdf) (code solution)

03/02: HW2 is up. Due Feb 15th. (pdf) (tex) (template code) (solution pdf) (code solution)

25/02: HW3 is up. Due Mar 6th. (pdf) (tex) (solution pdf)

25/02: Quiz 1 solution up. (pdf)

12/03: HW4 is up. (pdf)(tex) (Solution pdf)

15/03: Template code for loopy BP is up. (code) (Caution: These may be buggy/inefficient.)

31/03: Quiz 2 is scheduled on Apr 10, Tuesday.

31/03: HW5 is up. (pdf)(tex) (solution pdf)

16/04: HW6 is up. (pdf) (tex) (template code) (solution pdf) (code solution)

19/04: Template code for RBM and LDA are up. (RBM code) (LDA code) (Caution: These may be buggy/inefficient.)

25/04: Course wrapped up! Final exam on May 8th, 2PM.

05/05: Quiz 2 (Question paper) (Solution draft)

06/05: FINAL EXAM : CALCULATORS AND ONE A4 SHEET HANDWRITTEN CHEAT SHEET ALLOWED.

Class notes:

The class notes are in a VERY preliminary stage and should be taken only as a broad guideline. The real intent of the class notes is to just get a sense of what was covered in class and refer to the book/other references for details. When in doubt always consult the book or me in person. If the notes seem to contradict the book in any way, trust the book.

Useful links for practice problems apart from DKNF:

David Sontag's course page.

Devavrat Shah's course page.

Abbreviations:

BT: Bertsekas and Tsitsiklis. Introduction to probability.

DKNF: Daphne Koller and Nir Friedman: Probabilistic graphical models

CB: Christopher Bishop: Pattern recognition and Machine learning.