Learning to Plan via a Multi-Step Policy Regression Method

Stefan Wagner, Michael Janschek, Tobias Uelwer +1

We propose a new approach to increase inference performance in environments that require a specific sequence of actions in order to be solved. This is for example the case for maze…