30h Th, 5h Pr, 40h Proj.
Number of credits
Language(s) of instruction
Organisation and examination
Teaching in the first semester, review in January
Units courses prerequisite and corequisite
Prerequisite or corequisite units are presented within each program
Learning unit contents
Machine learning studies methods that automatically build a general solution to a problem from a set of solutions of specific instances of this problem. Its applications are multitudinous: extraction of medical diagnostic decision rules from clinical databases; bioinformatics; construction of credit allocation procedures from bank customer databases; computer vision; modeling, optimisation and control of complex systems; automatic syntesis of algorithms; extraction of knowledge from human experts...
The theoretical part of the course introduces the different types of machine learning problems (supervised learning: classification and regression, unsupervised learning: clustering and dimensionality reduction), the main underlying principles (bias/variance tradeoff, cross-validation, model selection) as well as the main families of methods (linear regression, k nearest neighbors, decision trees, ensemble methods, support vector machines, artificial neural networks, k-means...). Theoretical lectures are complemented with practical projects that allow the students to get acquainted with the theoretical concepts and the main method families by carrying out experiments on artificial and real datasets.
Learning outcomes of the learning unit
At the end of the course, the student will be able to analyze the theoretical (computational and statistical) properties of the most important machine learning algorithms, to apply them in practice, and to assess in a sound way their performances.
Prerequisite knowledge and skills
Elements of probability calculus, statistics, algorithmics, and optimization (as taught for example in the bachelor in engineering or in computer science).
Planned learning activities and teaching methods
Theoretical ex cathedra lectures combined with projects using the computer. Three projects are organized during the semester. The first two are intended to put into practice the course material by answering theoretical questions and carrying out experiments on artificial data sets. The third project is organized in the form of a student competition aimed at obtaining the best performances on a real supervised learning problem. The three projects are to be carried out by groups of typically two students.
Mode of delivery (face to face, distance learning, hybrid learning)
Theoretical lectures are face-to-face. Projects are carried out remotely.
Recommended or required readings
Slides of the theoretical lectures are available on the course website. Links to additional material (books or papers) are provided on the same web page and at the end of each set of slides.
Assessment methods and criteria
Exam(s) in session
Written work / report
The evaluation is based on the three projects (40%) and an oral exam (60%).
Each project will require to submit a written report and the source code of the solutions in due time. In addition, each group will be given the opportunity to present his solution to the third project (competition) in front of the class. The projects are inseparable from the teaching unit and considered as compulsory. Students who have not realised the projects and / or who have not submitted the expected reports within the prescribed time or in the prescribed form will not be allowed to take the exam.
The goal of the oral exam will be to assess the understanding of theoretical lectures. Students will have to present one or several parts of the course and answer questions covering the whole course material.
The lectures are given during the first semester, every Wednesday from 9:00am to 12:30am.
Web page: http://www.montefiore.ulg.ac.be/~lwh/AIA
Teachers:Louis Wehenkel (L.Wehenkel@uliege.be), Pierre Geurts (firstname.lastname@example.org)
Teaching assistants Yann Claes (email@example.com)