Machine Learning and Optimal Control
Organization
Credits: 5 ECTS
Term: Winter 22/23
Lecture & Exercise
Lecturer: Prof. Dr.-Ing. Timm Faulwasser
Schedule:
- Lecture: Wednesday 12:00 - 14:00 (SRG1-2.010)
- Exercise: Every other Wednesday 14:00 - 16:00 (CT-SR ZE 04)
Starting: 26.10.2022
Language: English
Examination: Oral exam
Office hours: On demand
Enrollment: Please sign up for the course on LSF. You will then be enrolled in Moodle automatically.
Majors:
M.Sc. Electrical Engineering and Information Technology
M.Sc. Automation and Robotics: Module
Content
Machine Learning (ML) is a key technology of the 21st century. Applications for ML in technical and communicational systems are already ubiquitous. This lecture provides a system and control theory-based introduction for different aspects of ML. Starting with the fundamental distinction between unsupervised, supervised, and reinforcement learning, the following topics will be covered:
- Reinforcement Learning with its connections to Optimal Control (see Hamilton-Jacobi-Bellman equation and Dynamic Programming) and Model Predictive Control (MPC)
- Formulation of discrete and continuous state-spaces
- Formulation of supervised Deep Learning as an Optimal Control Problem
- Data-driven approaches of MPC for linear systems
The applications to these ML approaches are formally analyzed and tested with standard software (e.g. Matlab or Python).