/old-reward-surfaces

Original version of reward surfaces.

Primary LanguagePython

Original RL surfaces

Idea: plot reward surfaces around trained models in RL baselines zoo

Repository contains copies of these models, so it is quite large.

To recreate experiments

The experiment data in the vis folder can be recreated with:

python generate_data.py

Note that this may take a long, long time, i.e. weeks, even with a powerful CPU.