tensowflow_ddpg

Deep Deterministic Policy Gradient

DDPG method is RL method that utilizes DPG Theorem, that can handle continuous action space. This implementation depends on Tensowflow and this RL model is experimented on aigym. Following papers are referred when implementing this model.

• https://arxiv.org/pdf/1509.02971.pdf
• https://arxiv.org/pdf/1511.04143.pdf

How to use

git clone https://github.com/sykwer/tensorflow_ddpg.git
cd tensowflow_ddpg
python main.py

Experiment

Experiment on InvertedPendulum-v2

(episode - total reward)

Resources

If you can conprehend Japanese article, refer to following articles! I have written several articles to understand DDPG theorem.

• https://sykwer.hatenablog.jp/entry/2018/03/08/105711
• https://sykwer.hatenablog.jp/entry/2018/03/10/Understand_DDPG_step_by_step_%282%29