With its simplicity, deep RL has the promise of being a driving force behind many AI breakthroughs. However, it often falls short of this promise due to the often unexpected complexity of the required training pipelines. This tutorial sheds light on the causes of this complexity and provides insights on how to deal with and mitigate them by means of AutoRL.
The purpose of this tutorial is to make RL accessible to the wider AI community and RL experts alike. In the first half, the tutorial will cover the mechanisms that make RL difficult to work with. In the latter half, the tutorial provides practical recommendations and an overview of available tools that help overcome this burden when aiming to use RL in practice.
