Introduction to Graphical Models

Guillaume Obozinski - Francis Bach
Ecole des Ponts, ParisTech - INRIA/ENS - INRIA/ENS

Master recherche specialite "Mathematiques Appliquees",
Parcours M2 Mathematiques, Vision et Apprentissage (ENS Cachan), 1er semestre, 2016/2017

This year, the class will be taught in English. Homeworks and project may be submitted in English or in French.

Classes will take place on Wednesdays from 9am to 12pm at ENS Cachan, in Amphi Curie.

Dates of classes

Below is the tentative schedule, with scribe notes from last year that might be updated for some lectures.

Link to the statistical review of the master, taking place on September 28rd and 29th. The material of the review is a prerequisite for this class.

Date	Lecturer	Topics	Corresponding chapters in class notes	Scribe notes
October 5th	Guillaume Obozinski	Introduction Maximum likelihood	5 Slides (intro) Slides ML	Huu Dien Khue Le, Robin Benesse lecture1.pdf lecture1.zip
October 12th	Guillaume Obozinski	Linear regression Logistic regression Generative classification (Fisher discriminant)	6, 7 Slides Regression	Aymeric Reshef, Claire Vernade lecture2.pdf lecture2.zip
October 19th	Guillaume Obozinski	K-means EM Gaussian mixtures	10, 11 Slides EM	Marie d'Autume, Jean-Baptiste Alayrac lecture3.pdf lecture3.zip
October 26th	Francis Bach	Graph theory Directed graphical models Undirected graphical models	2	Jaime Roquero, JieYing Wu lecture4.pdf lecture4.zip
November 2nd	Francis Bach	Exponential families Information theory	8, 19	Thomas Belhalfaoui, L&eacutena&iumlc Chizat lecture5.pdf lecture5.zip
November 9th	Francis Bach	Gaussian variables Factor analysis	13, 14	Lucas Plaetevoet, Ismael Belghiti lecture6.pdf lecture6.zip
November 16th	Guillaume Obozinski	Sum-product algorithm HMM	4, 12	Pauline Luc, Mathieu Andreux lecture7.pdf lecture7.tex
November 23rd	Guillaume Obozinski	Approximate inference Sampling Variational inference	21	Khalife Sammy, Maryan Morel lecture8.pdf lecture8.zip -- Basile Cl&eacutement, Nathan de Lara lecture9.pdf lecture9.tex
November 30th	Francis Bach	Bayesian methods Model selection	5.1 and 5.3	Gauthier Gidel, Lilian Besson lecture10.pdf lecture10.zip
December 14th (Amphi Curie)		Final Exam
January 4th Batiment Cournot (C102-103)		Project poster session

Homework (tentative schedule)

Homework 1, due October 31st, 2016 (on the Moodle): Homework | Data.

Projects

Updated page with current project proposals.

The final project allows a further understanding of certain aspects of the course. The following schedule has to be respected.

November	Choose a project and form teams of three.
Before 2016/11/18	Send an email to the two teachers, in which all members of your team are cc'ed to request our agreement on your choice of team and project topic.
Before 2016/12/09	Submit a progress report of 1 page including first results, on the Moodle.
On 2017/01/04	Poster session in Batiment Cournot (C102-103) - 9am to 12pm
Before 2016/01/11	Submit your project report (~6 pages, on the Moodle)

Description

This course provides a unifying introduction to probabilistic modelling through the framework of graphical models, together with their associated learning and inference algorithms.

References - Class notes

The course will be based on the book in preparation of Michael I. Jordan (UC Berkeley). Printed version of parts of the book (playing the role of the "polycopie") will be available from the Master's administrative assistant one or two weeks after the beginning of classes. We will notify you when they are ready for you to go and pick them up.

Last updated: November 8th, 2016.