Reinforcement Learning from Scratch

These are the exercises for the course "Reinforcement Learning from Scratch" that I give at Sioux Technologies.

Installation

If you want to run these exercises on your own computer you have to install the following packages.

Install an Anaconda environment like Miniconda (https://docs.conda.io/en/latest/miniconda.html) for Python 3.x.

Open an "Anaconda prompt" and type the following command to create a new python environment.

conda create -n rl_from_scratch python==3.11.* pytorch torchvision torchaudio -c pytorch

Or if you have an NVIDIA GPU:

conda create -n rl_from_scratch python==3.11.* pytorch torchvision torchaudio pytorch-cuda=11.8 -c pytorch -c nvidia

Install additional packages.

conda install ipython ipykernel swig
pip install stable_baselines3[extra]
pip install gymnasium[box2d]

Open this repository with Visual Studio Code and select rl_from_scratch as the Python environment.
Have fun

Name		Name	Last commit message	Last commit date
Latest commit History 15 Commits
data		data
figures		figures
trained		trained
.gitignore		.gitignore
Demo #1 - Value functions.ipynb		Demo #1 - Value functions.ipynb
Demo #2 - Policy iteration.ipynb		Demo #2 - Policy iteration.ipynb
Demo #3a - Taxi.ipynb		Demo #3a - Taxi.ipynb
Demo #3b - CartPole.ipynb		Demo #3b - CartPole.ipynb
Demo #3c - LunarLander.ipynb		Demo #3c - LunarLander.ipynb
Demo #3d - Pong.ipynb		Demo #3d - Pong.ipynb
Exercise #1 - Dynamic Programming.ipynb		Exercise #1 - Dynamic Programming.ipynb
Exercise #2 - Gymnasium.ipynb		Exercise #2 - Gymnasium.ipynb
Exercise #3 - Q-learning.ipynb		Exercise #3 - Q-learning.ipynb
Exercise #4 - Deep Q Network.ipynb		Exercise #4 - Deep Q Network.ipynb
Exercise #5a - Reinforce.ipynb		Exercise #5a - Reinforce.ipynb
Exercise #5b - Actor-Critic.ipynb		Exercise #5b - Actor-Critic.ipynb
Exercise #6 - Stable-Baselines3.ipynb		Exercise #6 - Stable-Baselines3.ipynb
estimator.py		estimator.py
mdp.py		mdp.py
readme.md		readme.md
utils.py		utils.py