2017-2018 / INFO8004-1

Advanced Machine learning

Durée

30h Th, 5h Pr, 45h Proj.

Nombre de crédits

 Master en science des données, à finalité5 crédits 
 Master : ingénieur civil en informatique, à finalité5 crédits 
 Master : ingénieur civil en science des données, à finalité5 crédits 
 Master en sciences informatiques, à finalité5 crédits 

Enseignant

Pierre Geurts, Louis Wehenkel

Langue(s) de l'unité d'enseignement

Langue anglaise

Organisation et évaluation

Enseignement au deuxième quadrimestre

Unités d'enseignement prérequises et corequises

Les unités prérequises ou corequises sont présentées au sein de chaque programme

Contenus de l'unité d'enseignement

This course focuses on advanced supervised and unsupervised machine learning methods, as well as on non-standard learning protocols, such as semi-supervised learning, transfer learning, and active learning.
 
From the practical viewpoint, the emphasis will be on machine learning problems dealing with high-dimensional and structured representation spaces (e.g. texts, images, videos, time-series, and graphs). In terms of methods, the focus will be on advanced method families such as deep neural networks, Gaussian processes, or kernel-based methods.
 
The course will also allow the students to become familiar with the main theoretical frameworks used to analyze machine learning methods.

Acquis d'apprentissage (objectifs d'apprentissage) de l'unité d'enseignement

At the end of the class, the students are expected to master the state of the art in the field of machine learning. They will be able to implement, combine, or extend existing algorithms to address very complex machine learning tasks, and they will have the theoretical background to read scientific papers and start doing research in the field.

Savoirs et compétences prérequis

The course relies strongly on linear algebra, probability calculus, elementary statistics, and notions of optimization, as well as good knowledge of data structures and algorithms.
A prior introduction to machine learning, information theory, and graphical models, is useful but not mandatory.
A strong interest in advanced applications of machine learning is expected from the students, as well as willingness to self-learn in an autonomous way and to present their ideas in a clear fashion during the course lectures.

Activités d'apprentissage prévues et méthodes d'enseignement

This course is organized in full-English mode.
Ex-cathedra lectures given by the professors will be supplemented by discussion sessions with the students around key papers in the field, and by research seminars given by external speakers.
Personal student projects will consist in the use of machine learning toolboxes to implement and evaluate advanced algorithms on selected problems, and in critical reading and discussion of a selection of scientific papers on the topics related to the course.

Mode d'enseignement (présentiel ; enseignement à distance)

face-to-face

Lectures recommandées ou obligatoires et notes de cours

See course web-page:  http://www.montefiore.ulg.ac.be/~geurts/AML.html
 

Modalités d'évaluation et critères

The students will carry out two mandatory assignments by groups of 4-6 students. The first assignment will consist in reading a recent paper and presenting it during a lecture to the rest of the students. The second assignment will consist in implementing an advanced machine learning application, and presenting the choices made to do so to the other students.
These two assignments will intervene each one for 30% of the final grade. An oral exam will be organized, intervening for the remaining 40% of the grade.

Stage(s)

None

Remarques organisationnelles

This course is organized in full-English mode.
 

Contacts

L.Wehenkel@ulg.ac.be and P.Geurts@ulg.ac.be