This paper provide the reader with the conceptual tools needed to get started on research on offline reinforcement learning algorithms:
reinforcement learning algorithms that utilize previously collected data, without additional online data collection.

However, the limitations of current algorithms make this difficult. We need to understand these challenges, and using modern deep reinforcement learning methods.

Also, this paper describes some potential solutions that have been explored in recent work to mitigate these challenges, along with recent applications.

Powerpoint for this talk

Powerpoint for this talk