392143 Applied Cognitive Computing: Advancing Reinforcement Learning Agents through Cognitive Mechanisms (S) (SoSe 2026)

This seminar explores the topic of reinforcement learning agents and how they can be optimised using human cognitive mechanisms such as abstraction, intrinsic motivation, mental simulation, few-shot problem solving, transfer learning or goal-directed planning. We will start with the field of Reinforcement Learning and then move on to research on cognitive abilities. Finally, research combining these two fields will be discussed. Each student or pairs of students (depending on the number of participants) has to present (~15 min.) a research paper in the seminar. In addition, the exercises will be used to work on a project where an implementation of cognitive mechanisms in RL is carried out, as well as a report on the implementation including a theoretical background. The presentation, implementation and report are part of the “Prüfungsleistung” of the seminar and the exercise. Some programming experience in Python is recommended.