Reinforcement Learning
The aim of the course is to introduce basic as well as modern concepts of reinforcement learning. This includes Markov decision processes, dynamic programming, model-free prediction and control, temporal difference learning, function approximation methods, policy gradient and actor-critic methods, and modern applications of reinforcement learning.
The course consists of two modules: theory (3.5 credits) and project (4 credits).
Teaching Format
Instruction is given in the form of lectures, exercise sessions and supervision.
Assessment
The course is assessed through a written exam, and project assignment.
Both parts of the course are graded on a scale from A to F, where A to E are passing grades. To complete the course, a passing grade is required on both parts, and the final grade of the course is determined by weighing the grades from the course modules, where each grade is weighed in relation to the scope of the course module.
Examiner
A list of examiners can be found on
"Reinforcement Learning: An introduction", R.S. Sutton and A.G. Barto, 2nd Edition, MIT Press, Cambridge, MA (2018) (e-book available via the university library)
"An Introduction to Deep Reinforcement Learning", V. François-Lavet, P. Henderson, R. Islam, M.G. Bellemare, J. Pineau, Foundations and Trends in Machine Learning: Vol. 11: No. 3-4, pp 219-354 (2018) (available for download at arxiv.org)
New student
During your studies
Course web
We do not use Athena, you can find our course webpages on kurser.math.su.se.
