myrl

Implementations of some standard reinforcement learning algorithms

• Written in pytorch
• All implentations are encapsulated into a single file
• Algorithms implemted as classes which take OpenAI gym env as inputs
• All classes have uniform interface with train and eval methods
• All references are given in the code

Algorithms -

• Vanilla Policy Gradient (VPG)
• Deep Q Network (DQN)
• Deep Deterministic Policy Gradient (DDPG)
• Twin Delayed DDPG (TD3)
• Soft Actor Critic
• Proximal Policy Optimization
• Trust Region Policy Optimization

Features to add -

• Better hyperparamter management
• Adding tensorboard logging
• Add documentation for each algorithm
• Add results for each algorithm