Reinforcement Learning in R

Markus Dumke
Reinforcement learning is a class of algorithms to solve sequential decision making problems. While there are many implementations available in Python, there are nearly no algorithms available in R. This thesis introduces the reinforcelearn package, which aims to make a range of important reinforcement learning algorithms available for R users. The first section describes the reinforcement learning problem and the corresponding notation, the second section explains algorithms to solve
more » ... ent learning problems, the third section then introduces the R package reinforcelearn and how to use it. Furthermore we introduce Double Q(σ) and Q(σ, λ), two new reinforcement learning algorithms which subsume Q-Learning, Sarsa and Q(σ).
doi:10.5282/ubm/epub.59093 fatcat:5rlt6txchvhkpjooyxlpuaye4m