Python keras + tensorflow implementation of DDPG solving modified open gymAI pendulum-v0 environment
Corrected and modified implementation from Reinforcement Learning w/ Keras + OpenAI: Actor-Critic Models of DDPG using keras + tensorflow.
It solves open gymAI pendulum-v0 environment
About METHOD: Continuous control with deep reinforcement learning
python ac_pendulum.py
It shows animation every 5 episodes.
It reaches desired behaviour after around 40-60 episodes.