RL Algorithms

A collection of implementations of RL algorithms in Python. Developed for my own personal learning as I work through papers and tutorials.

Algorithms implemented

1. Simple Policy gradient:

• using only a policy network and no advantage function
• also implemented using reward-to-go

2. Vanilla Policy Gradient

• using reward-to-go, simple advantage function (Q(s, a) - V(s)) and GAE

3. Deep Q-network with experience replay

• Based off of the original DQN paper (Mnih et al (2013))

4. Synchronous Actor Critic (A2C)

Resources used

• OpenAI Spinning Up