nl en

Seminar Advanced Deep Reinforcement Learning


Admission requirements

Recommended prior knowledge

It is strongly recommended that students have successfully completed:
1. Reinforcement Learning
2. Introduction to Deep Learning
Students must be highly interested in scientific research and proficient in programming and machine learning and writing reports. This is an advanced course. We will work intensively on state of the art research topics.


Deep Reinforcement learning is a field of Artificial Intelligence that has attracted much attention, after history making achievements in Games and Robotics.
In this seminar course current scientific topics in recent Deep Reinforcement Learning research papers will be studied. Topics can include · Transfer Learning · Meta Learning · Model-based Reinforcement Learning · Sample Efficiency · Hierarchical Reinforcement Learning

General application areas are games, robotics, and other autonomous behavior. Typical benchmark applications of deep reinforcement learning are OpenAI Gym, DeepMind Bsuite, ALE, OpenSpiel, Meta World, PolyGames, and others.
Capacity is limited, attendance is mandatory, level of enthusiasm & difficulty is high. We will work in small groups studying recent scientific papers on the above topics.

This is a “third semester” course, after you have followed master courses on Reinforcement Learning and on Deep Learning.
Please see the Course Website

Course objectives

We study some of the latest research papers in advanced topics in deep reinforcement learning. Students learn about the latest research, by reading, understanding, implementing and presenting recent scientific insights in Deep Reinforcement Learning.
A paper will be chosen, the student will (1) reimplement (part of) the work, (2) present their work, and (3) write a paper about it.


The most recent timetable can be found at the Computer Science (MSc) student website.
Detailed table of contents can be found in Brightspace.

Mode of instruction

Student presentations about a research topic & implementation (with peer feedback) and papers (with feedback from lecturers). Together we study recent literature on topics in a common theme within the field of Deep Reinforcement Learning.

Course load

Hours of study: 168:00 hrs (= 6 EC)
Lectures: 26 hrs
At home preparation: 142 hrs

Assessment method

The final grade is determined by: · Presentations · Active participation · A written report/paper · Peer review of a programming implementation

The teacher will inform the students how the inspection of and follow-up discussion of the exams will take place.

Reading list

The reading list is changing every academic year according to the subject of study. See the course page of Brightspace for more information.


You have to sign up for courses in uSis. Check this link for information about how to register for courses.
Important: due to the format of the course, there is a limit on the number of participants: at most 12 students can participate in this course.

Register by 1) signing up in uSis and 2) sending an email to the lecturers before 1 August with your grades for Reinforcement Learning and Deep Learning.

Students with high grades for Reinforcement Learning and Deep Learning (Introduction to -, or - & Neural Networks) (unweighted average of the two courses) are invited by the lecturers for an intake meeting in August.


Lecturer: Prof.dr. Aske Plaat
Website: Brightspace
Course Website: here