We include a number of interactive jupyter notebooks demonstrating various feature of the library, as well as reproducing known results.
-
The tutorial notebook includes a basic usage tutorial of both environment types. It also demonstrates how to generate figures displaying learned representations from agents, such as value or successor maps.
-
The Grid experiments notebook includes code to evaluate agents on a set of grid environment tasks inspired by Russek et al., 2017.
-
The graph experiments notebooks: (Experiment 1) and (Experiment 2) include code to evaluate agents on graph tasks taken from Momennejad et al., 2017.