| Detailed information |
| Original study plan |
Bachelor's programme Artificial Intelligence 2025W |
| Learning Outcomes |
Competences |
| Students have a basic understanding of core concepts, theories, and methods relating to the field of reinforcement learning. They understand what kinds of problems can be suitably modeled as sequential decision processes and addressed with reinforcement learning algorithms.
|
|
| Skills |
Knowledge |
| Students understand the basic concepts of and assumptions behind Markov decision processes (k2), and how to model a given sequential decision and optimisation task as a reinforcement learning problem (k3). They know fundamental algorithms of reinforcement learning, and can set up reinforcement learning agents and experiments (k3/4).
|
Fundamental concepts of reinforcement learning, underlying modeling assumpsions, and basic algorithms of reinforcement learning:
- Solution methods for k-armed Bandits, and their practical application
- Formal treatment of Markov Decision Problems (MDPs)
- Theory for solving MDPs
- Table-based solution methods for MDPs with discrete state spaces
- Selected approximate solution methods for MDPs with continuous state spaces
- Outlook on approximate solution of MDPs with very large discrete state spaces and deterministic dynamics (e.g. board games such as „Chess“ and „Go“)
|
|
| Criteria for evaluation |
Positive grade on the final exam (written).
|
| Methods |
Standard lectures. Positive reinforcement of active lecture participation through rewards in the form of small chocolate treats
|
| Language |
English |
| Study material |
Richard S. Sutton and Andrew G. Barto. 2018. Introduction to Reinforcement Learning (2nd. edition). MIT Press, Cambridge, MA, USA.
|
| Changing subject? |
No |
| Further information |
This lecture course (VO) and the corresponding exercise course (UE) form a didactic unit. The study results described here are achieved through the combination of these two courses.
|
| Corresponding lecture |
in collaboration with 536MLPEREIU20: UE Reinforcement Learning (1.5 ECTS) equivalent to
536MLPEREIK19: KV Reinforcement Learning (4.5 ECTS)
|
