General information
| Course name | Lecture: 39734 Approximate Dynamic Programming (Reinforcement Learning) |
| Subtitle | vormals "Advanced Topics in Management Science" |
| Course number | 39734 |
| Semester | WiSe 24/25 |
| Current number of participants | 120 |
| expected number of participants | 200 |
| Home institute | Lehrstuhl für Betriebswirtschaftslehre mit Schwerpunkt Management Science / Operations and Supply Chain Management |
| Courses type | Lecture in category Lehre (mit Prüfung) |
| Next date | Monday, 20.01.2025 14:00 - 16:00 Uhr, Room: (WIWI) HS 7 |
| Type/Form | |
| Participants |
BA, WI, CMT, M, INF, AIE |
| Pre-requisites |
Mathematical maturity and the ability to write down precise and rigorous arguments. Solid basic knowledge of modeling and optimization. |
| Learning organisation |
• Block course with interactive elements and classroom discussions; • Solution and discussions of exercises and case studies; • Online forums and discussions; • A take-home mock exam to simulate the final exam of the course. Discussion of this mock exam; • Blended learning, such as usage of software examples, videos and web-based exercises |
| Performance record |
a) Final exam 100 % or b) Final exam 90% + 10 % for completing optional assignments during the semester (with reservations) |
| SWS |
2 |
| Literatur |
Bertsekas, D. P., and Tsitsiklis, J. N. (1996). Neuro-Dynamic Programming. Athena Scientific: Massachustetts. Bertsekas, D. P. Dynamic Programming and Optimal Control. Athena Scientific: Massachustetts. Bertsekas, D. P. Dynamic Programming and Optimal Control: Approximate Dynamic Programming. Athena Scientific: Massachustetts. Bertsekas, D. P. Abstract Dynamic Programming. Athena Scientific: Massachustetts. Powell, W. B. Approximate Dynamic Programming. John Wiley and Sons. |
| Qualifikationsziele |
After successful participation in the module, students will be able to: Represent deterministic and stochastic optimization problems as dynamic programs Solve deterministic and stochastic optimization problems exactly, incl. with the backward induction, value and policy iteration methods Understand foundations of Markov chains and Markov decision processes and meaningfully apply them to solve stochastic optimization problems Apply the approximate dynamic programming (reinforcement learning) algorithm, critically appreciate variations in its design Critically evaluate the potential of AI and further the generic heuristic solution approaches in the light of the recent successes and of the no-free-lunch |
| Workload |
Lecture 2 SWS (30 h attendance and 45h own work) Excercise 2 SWS (30 h attendance and 45 own work) Calculation basis: 15 weeks in a semester, including an examination week; each SWS corresponds to 60 minutes |
| Miscellanea |
Sofern die Klausur PN 266193 "Advanced Topics in Management Science" bereits erfolgreich absolviert wurde, ist es nicht möglich, zusätzlich die Klausur PN 266194 "Approximate Dynamic Programming (Reinforcement Learning)" zu absolvieren. |
| ECTS points |
5 |
