Course: Probabilistic Graphical Models

Overview

Formalities:

• Half semester course, 3 ECTS
• please register in iQ

• first meeting: Tuesday, November 29 (8:45!, because of a scheduling conflict)
• starting on Thursday, December 1, the course will take place at its regular time slots
• starting on Thursday, December 1, the course will be in Mondi 2 on Thursdays and Mondi 3 on Tuesdays
• the lecture and recitation on December 6 will take place (originally, that wasn't clear because of travel)
• there will be no lecture on December 8 because of the public holiday
• on January 10th, there will be a guest lecture by Sesh Kumar on approximate inference
• on January 24th, there will be a guest lecture by Vladimir Kolmogorov on energy minimization using graph cuts an related techniques
• Tuesday, January 24th, will be the last lecture. There will be final recitation afterwards, but no lecture on Thursday.

• exercise sheet 1: PDF (discussed at the recitation on Dec 6)
• Note: typos in 2d) and 2g) were fixed on 29/11/16 14:00 and 15:25
• exercise sheet 2: PDF (discussed at the recitation on Dec 13)
• exercise sheet 3: PDF (discussed at the recitation on Dec 20)
• exercise sheet 4: PDF (discussed at the recitation on Jan 10)
• exercise sheet 5: PDF (discussed at the recitation on Jan 17)
• exercise sheet 6: PDF (tentative, beware of typos; to be discussed at the recitation on Jan 24)

Course Materials

• presentation slides will be posted here after each lecture
• Lecture 1: Intro and Refresher of Probabilities PDF
• Lecture 2: Bayesian Networks (directed graphical models) PDF
• Lecture 3a: Excurse Causality PDF
• Lecture 3b: Markov Networks (undirected graphical models) PDF
• Lecture 4: Factor graphs, Parameter Estimation PDF
• Lecture 5: Maximum-A-Posteriori Estimation, Bayesian Inference, EM PDF
• Lecture 6: Inference and Learning in Hidden Markov Models PDF
• Lecture 7: Maximum Entropy Models, (Loopy) Belief Propagation PDF
• Lecture 8 (Sesh Kumar): Approximate Inference
• Lecture 9: Conditional Random Field Learning PDF
• Lecture 10: Energy Minimization, Structured Loss Functions PDF
• Lecture 11: Structured Support Vector Machines PDF
• Lecture 12: (Vladimir Kolmogorov): Graph cuts
• textbook: David Barber, "Bayesian Reasoning and Machine Learning", Cambridge University Press, 2011, ISBN-13: 978-0521518147
• more exhaustive textbook: Koller, Friedman, "Probabilistic Graphical Models: Principles and Techniques", The MIT Press, 2009, ISBN-13: 978-0262013192