Information on individual educational components (ECTS-Course descriptions) per semester

Basics in machine learning, deep-learning, statistics, probability theory and linear algebra. 

Reinforcement Learning is an area of machine learning concerned with how "intelligent agents" can learn a strategy to reach a certain goal. In recent years, such agents (based on deep-learning) were able to beat human top players in Go, Starcraft II and Dota in a fair setting. This achievement was thought to be impossible only a few years ago. Methods from reinforcement learning however can not only be used in a game setting but provide various different possible applications from cooperative robots to autonomous vehicles. In this course the principles of reinforcement learning are covered:

• the reinforcement learning problem
• basics (action-value methods, markov decision process, dynamic programming, monte carlo methods, temporal difference learning)
• Q-learning, Deep-Q-Network (DQN)
• Policy Optimization, Asynchronous Actor-Critic (A2C)
• applications

 Examples and projects will be implemented in python using frameworks such as scikit-learn, numpy and keras (tensorflow).

The students are able to

• describe the difference between supervised-learning and reinforcement-learning
• explain the reinforcement-learning problem
• recognize for which type of problems methods from reinforcement-learning are suited
• build, train and employ reinforcement-learning models
• use and adapt state-of-the-art models based on keras and tensorflow

Lectures and mini-projects in groups. 

Written Exam: 75% (must be positive)

Excersises: 25% (must be positive)

For a positive grade, a minimum of 50% of the possible points must be achieved in each part of the examination.

 None

• Goodfellow, Ian; Bengio, Yoshua; Courville, Aaron (2016): Deep Learning. MIT Press. Available at: URL: https://www.deeplearningbook.org/
• Mitchell, Tom (2017): Machine Learning. Available at: URL: http://www.cs.cmu.edu/~tom/NewChapters.html
• Norvig, Peter; Russel, Stuart (2021): Artificial Intelligence: A Modern Approach, Global Edition.
• Sutton, Richard; Barto, Andrew (2002): Reinforcement Learning. MIT Press.

Face-to-Face with selected online elements.