Reinforcement Learning

This is a repo containing explanations and implementations of the following Reinforcement Learning algorithms. Note that, if you want to see the notebooks clearly (with LaTeX and stuff) you need to clone the repo.

• git clone [email protected]:ceteke/RL.git
• cd RL
• jupyter notebook

Tabular

The tabular methods are demonstrated on OpenAI Gym's FrozenLake environment.

Dynamic Programming

• Policy Iteration
• Value Iteration

Monte Carlo

• On-Policy Control
• Exploring Starts
• Off-Policy Control

Temporal Difference

• On-Policy TD (Sarsa)
• Off-Policy TD (Q-Learning)
• n-Step On-Policy TD (Sarsa)

Approximation

• Linear Sarsa
• Average Reward Linear Sarsa
• n-Step Linear Sarsa
• Actor-Critic Action Value Policy Gradient

Policy Gradient

• REINFORCE (Monte Carlo Policy Gradient)
• Hindsight Experience Replay

Deep Reinforcement Learning