/DDPG

Implementation of DDPG (Modified from the work of Patrick Emami) - Tensorflow (no TFLearn dependency), Ornstein Uhlenbeck noise function, reward discounting, works on discrete & continuous action spaces

Primary LanguagePython

Implementation of DDPG - Deep Deterministic Policy Gradient

Modified from the work of Patrick Emami: Deep Deterministic Policy Gradients in TensorFlow

Algorithm and hyperparameter details can be found here: "Continuous control with deep reinforcement learning" - TP Lillicrap, JJ Hunt et al., 2015

Tested on CartPole & Pendulum

Requirements

Gym and TensorFlow.

Modifications

  • Removed TFLearn dependency
  • Added Ornstein Uhlenbeck noise function
  • Added reward discounting
  • Works with discrete and continuous action spaces