DDPG

Continuous control with deep reinforcement learning

http://arxiv.org/abs/1509.02971

The goal of these algorithms is to perform policy iteration by alternatively performing policy evaluation on the current policy with Q-learning, and then improving upon the current policy by following the policy gradient.

TODO

• Batch Normalization
• Prioritized Experience Replay (https://arxiv.org/abs/1511.05952): to replay important transitions from reply Memory more frequently

Reference

• Continuous control with Deep Reinforcement Learning (by TP Lillicrap)
• Deterministic Policy Gradients (by TP D. Silver )
• Human-level control through deep reinforcement learning (by V. Mnih )