4SkyNet / learning2run

Our NIPS 2017: Learning to Run approach

Our NIPS 2017: Learning to Run approach

Where "our" refers to: Adam Stelmaszczyk, Piotr Jarosik

Write up: https://medium.com/@stelmaszczykadam/our-nips-2017-learning-to-run-approach-b80a295d3bb5

Main files:

• run_osim.py to run baselines PPO
• baselines/baselines/pposgd/pposgd_simple.py observation processing for PPO
• example.py to run keras-rl DDPG (with old observation processing)
• es/localhost/launch.py to run Evolution Strategies
• osim-rl/osim/env/run.py#L67 reward hacking