This repository ccontains a bunch of notes and code that I have implemented through self-learning. Most of these resources are either through a free course online or through a textbook that matched the content that I'm interested in.
- Policy Iteration using iterative policy evaluation on FrozenLake-v1
- Value Iteration on FrozenLake-v1
- First-visit MC Prediciton on FrozenLake-v1
- Monte Carlo Exploring Starts on FrozenLake-v1
- REINFORCE: Monte-Carlo Policy-Gradient Control on MountainCar-v0
- REINFORCE with Baseline: Monte-Carlo Policy-Gradient Control on MountainCar-v0
- Tabular TD(0) on FrozenLake-v1
- Sarsa on FrozenLake-v1
- Q Learning on FrozenLake-v1
- Double Q Learning on FrozenLake-v1
- N-Step TD Learning on FrozenLake-v1
- N-Steo Sarsa Learning on FrozenLake-v1
- Episodic Semi-Gradient Sarsa on MountainCar-v0
- One-Step Actor-Critic on MountainCar-v0
FrozenLake is a "game" environment from Gym. This is part of there Toy Text environments, in which they have multiple other environments such as MuJoCo, Atari, Classic Control, etc. There is not a lot of detailed documentation and example, but this is from there official website.
MountainCar is a "game" environment from Gym. This is part of there Classic Control environments, in which they have multiple other environments such as MuJoCo, Atari, Classic Control, etc. There is not a lot of detailed documentation and example, but this is from there official website.
To run the code on your own computer, you need to install GymLibrary. There really isn't any good tutorial on how to run this except for this one that I found on here. I'm running all of the code on conda version 24.5.0 with python version 3.9.19.
The psuedo-code for all of the RL Code is from An introduction to Reinforcement Learning, Sutton and Barto, Second Edition