Valid from: Spring 2021
Decided by: Professor Thomas Johansson
Date of establishment: 2021-02-18
Division: Automatic Control
Course type: Third-cycle course
Teaching language: English
To teach the participants basic topics in deep reinforcement learning
Knowledge and Understanding
For a passing grade the doctoral student must understand advanced topics in reinforcement learning and implementation aspects of deep reinforcement learning using neural networks.
Competences and Skills
For a passing grade the doctoral student must be able to implement different deep reinforcement learning techniques based on Neural Networks.
Judgement and Approach
For a passing grade the doctoral student must
Q functions, Advanced Policy Gradients, Model Based planning, Model Based RL, Model Based Policy Learning, Control as Inference, Inverse Reinforcement Learning, Distributed RL, Challenges and Open problems
Sutton, Richard S. & Barto, Andrew G.: Reinforcement Learning, Second Edition. ISBN 9780262039246.
Types of instruction: Seminars, exercises, self-study literature review, miscellaneous. based on Berkely deep RL course "CS285" and a few assignments available on Github
Examination format: Miscellaneous.
Active participation in the course
Grading scale: Failed, pass
Admission requirements: Admitted to PhD program at an engineering faculty
Assumed prior knowledge: Basic reinforcement learning knowledge, basic neural network construction and training using tensorflow or pytorch
Selection criteria: None
Web page: http://www.control.lth.se/education/doctorate-program/study-circle-in-deep-reinforcement-learning/