The goal of this library is to provide and out of the box passive reinforcement learning algorithms that are easy to use. The work on your part is to define your problem as a Graph and the library takes care of the rest. The use of a graph, rather than the classic MDP formulation (S,A,R,P
), is to provide an easy/efficient way to define MDPs where actions vary across states.
With pip:
python -m pip install git+https://github.com/bi3mer/GDM.git@main
With pipenv
pipenv install -e git+https://github.com/bi3mer/GDM.git@main#egg=GDM
In addition to basic unit testing, tests/ provides and tests three different grid world environments, the gambler's ruin environment, and tic-tac-toe.