Deep RL

My implementations of popular deep reinforcement algorithms. Each algorithm is implemented in a single file for readability and ease of understanding.

Algorithms Implemented

Algorithm	Action Space	Implementation
Deep Q-learning (DQN)	Discrete	dqn.py
REINFORCE (with baseline)	Discrete	reinforce.py
Deep Deterministic Policy Gradient (DDPG)	Continuous	ddpg.py
Twin Delayed Deep Deterministic Policy Gradient (TD3)	Continuous	td3.py

Algorithms in Progress

• Proximal Policy Optimization (PPO)

Experiments

DQN vs. REINFORCE with / without baseline on Cart Pole

DDPG vs. TD3 on Half Cheetah

Videos of TD3 on Half Cheetah

5K timesteps into training (0.05% completed)

td3-halfcheetah-timestep-30k.mp4

205K timesteps into training (20.5% completed)

td3-halfcheetah-timestep-230k.mp4

955K timesteps into training (95.5% completed)

td3-halfcheetah-timestep-980k.mp4