This repository contains a set of implementations and functionality of purely online tabular reinforcement learning algorithms.
Currently we support Python 3.8.
You should be able to install table-rl from PyPi.
pip3 install table-rl
You can also install table-rl from source.
git clone https://github.com/prabhatnagarajan/table-rl.git
cd table-rl
pip3 install -e .
Reward functions, value functions, transition functions are all numpy arrays and matrices. We follow the conventions from Sutton and Barto's RL Book.
We focus only on online algorithms that interact with and get experience from the environment in a purely online way.
Use pytest on the test directories.
python -m pytest tests/