Reinforcement learning has gained considerable traction as it mines real experiences with the help of trial-and-error learning to model decision-making. Thus, this approach attempts to imitate the fundamental method used by humans of learning optimal behavior without the requirement of an explicit model of the environment. In contrast to many other approaches from the domain of machine learning, reinforcement learning works well with learning tasks of arbitrary length and can be used to learn complex strategies for many scenarios, such as robotics and game playing.

Our slide deck is positioned at the intersection of teaching the basic idea of reinforcement learning and providing practical insights into R. While existing packages, such as MDPtoolbox, are well suited to tasks that can be formulated as a Markov decision process, we also provide practical guidance regarding how to set up reinforcement learning in more vague environments. Therefore, each algorithm comes with an easy-to-understand explanation of how to use it in R.

We hope that the slide deck enables practitioners to quickly adopt reinforcement learning for their applications in R. Moreover, the materials might lay the groundwork for courses on human decision-making and machine learning.

Download the slides here

Download the exercise sheet here (solutions are available on request)