tianhaofu / p2il Goto Github PK

Implementation for Proximal Point Imitation Learning

The expert trajectories used during the experiments can be found here: https://drive.google.com/drive/folders/1ACP1QIqJ4KjBvwSqsyPEK5BoclueWCyx?usp=sharing

Place the folder LinearMDPs in the folder linear_PPIL/expert_trajs/ and the folder MDPs in the folders offline_PPIL/experts/ and deep_PPIL/experts/

The folder linear_PPIL contains the code for reproducing the experiments in Linear MDPs. The folder deep_PPIL contains instead the code for experiments with neural network function approximation.

Finally, the folder offline_PPIL contains the code for the offline version.