Duration
25h Th, 10h Pr, 45h Proj.
Number of credits
Lecturer
Language(s) of instruction
English language
Organisation and examination
Teaching in the second semester
Schedule
Units courses prerequisite and corequisite
Prerequisite or corequisite units are presented within each program
Learning unit contents
In a world where intelligent systems are increasingly autonomous, reinforcement learning (RL) is revolutionising decision-making across a range of complex problems (e.g., control of anti-UAV robots on a battlefield). From optimising robotic controls to developing strategies for financial markets, RL enables agents to learn from interactions with their environements and make decisions that maximise long-term rewards.
This course provides a comprehensive introduction to RL, focusing on both theoretical foundations and practical applications. As an example of theoretical thematics, we can mention learning in low-data environments (which is particularly useful for designing efficient medical treatments for chronic diseases such as for example obesity, alcoholism and cancer), operating in partially observable settings (problems met for example in robotics, in games or when interacting with energy markets) and coordinating multiple agents, a thematic that becomes increasingly important with the defense industry currently developing drone-swarm technologies. Practical applications of RL to real-world problems will include robotics, large language models (LLMs) and infrastructure management planning. Students will gain hands-on experience by implementing, training, and evaluating RL algorithms, preparing them to tackle cutting-edge challenges in various domains. The class will be organised along several plenary lectures, whose subjects include the following:
- RL Fundamentals
- Markov Decision Processes (MDPs)
- Deep Reinforcement Learning
- Low-Data RL
- Policy Gradient Methods
- Partially Observable Environment RL
- Multi-Agent Reinforcement Learning (MARL)
- Robotic RL
- LLMs in RL
Learning outcomes of the learning unit
At the end of the course, the student will be (i) familiar with a broad range of techniques for solving reinforcement learning problems, (ii) able to apply these techniques in practice and understand their key characteristics, (iii) able to effectively read and understand the scientific literature dedicated to reinforcement learning.
This course contributes to the learning outcomes I.2, II.1, II.2, II.3, III.1, IV.1, IV.3, VI.1, VI.2, VI.3, VII.2, VII.5 of the MSc in electrical engineering.
This course contributes to the learning outcomes I.2, II.1, II.2, II.3, III.1, IV.1, VI.1, VI.2, VI.3, VII.2, VII.5 of the MSc in computer science and engineering.
Prerequisite knowledge and skills
Basic knowledge in system theory, statistics, optimisation and machine learning.
Good coding skills are required.
It is also recommended students take the course INFO8010-1 Deep Learning or have completed an equivalent course.
Planned learning activities and teaching methods
The classes will include different elements: theoretical courses, analyses of scientific articles and exercises. Part of the theoretical material will be taught through inverse teaching.
Students will also have to work throughout the year on projects designed to implement the methodologies learned during the year on fairly simple examples.
Mode of delivery (face to face, distance learning, hybrid learning)
Face-to-face learning
Course materials and recommended or required readings
Other site(s) used for course materials
- Site Web Damien Ernst (http://www.damien-ernst.be)
Further information:
The teaching material is accessible on the class website, see: https://damien-ernst.be/teaching/
Exam(s) in session
Any session
- In-person
oral exam
Written work / report
Further information:
Any session :
- In-person
oral exam
- Remote
written work
- If evaluation in "hybrid"
preferred in-person
Additional information:
The evaluation consists of two parts: 60% for the oral examination at the end of the year and 40% for the projects. For the second session, the final grade will be computed the same way and the students will have the possibility to resubmit their projects if they want to get better grades for them.
Work placement(s)
Possibility for motivated students to do a (eventually paid) research internship on various topic (defence, energy,...) in RL.