Implementing algorithm from
Lillicrap, Timothy P., Jonathan J. Hunt, Alexander Pritzel, Nicolas Heess, Tom Erez, Yuval Tassa, David Silver, and Daan Wierstra. "Continuous control with deep reinforcement learning." arXiv preprint arXiv:1509.02971 (2015). https://arxiv.org/abs/1509.02971
Modify and/or run ddpg/trainer.py
Dependencies: Tensorflow r2.0, numpy, matplotlib, gym
- noiseless evals every X training episodes
- parametrize network architecture
- better hyperparams for Pendulum
- more tasks
From 10 runs of ddpg/trainer.py as is
