Deep-Reinforcement-Learning-in-Large-Discrete-Action-Spaces

Link to paper

Implementation of the algorithm in Python 3, TensorFlow and OpenAI Gym.

This paper introduces Wolpertinger training algorithm that extends the Deep Deterministic Policy Gradient training algorithm introduced in this paper.

I used and extended stevenpjg's implementation of DDPG algorithm found here licensed under the MIT license.

Master is currently only for continuous action spaces.

The branch discrete-and-continuous provides the ability to use the discrete environments of the gym.