AlphaGo knock off project for the course AI-programming at NTNU Trondheim
This is a reinforcement learning project that uses a on-policy neural network to learn how to play an intelligant game of Hex and Nim. The neural network is trained using a monte carlo tree search algorithm. The neural network is implemented using the PyTorch and the monte carlo tree search algorithm is implemented from scratch.
Models are saved during training and can be set against each other in a round robin tournament.
pip3 install -r requirements.txt
Set the desired configuration in config.yml
and run the program with the following command:
python3 main.py
The tests test if the monte carlo tree search algorithm is working as intended.
python3 tests.py