Dark Hex is the imperfect information version of the game Hex. This game has a really scarce work on it. In this repository I am including an implementation of the game along with some algorithms implemented specifically for DarkHex. Also I included the results of some of the experiments.
For details on DarkHex and any of the algorithms used as well as the experiments please check my thesis -link-.
- pONE
- Vanilla CFR
- FSI-CFR
- Backward Induction Best Response
- AlphaZero Approximate Best Response
- CFR+
- MCCFR
- NFSP
- Deep CFR
The library is uploaded on Pypi and can be installed using pip. Before pip called make sure you have the following packages installed:
They are mostly required for graph drawing and setup reasons. Please make sure you have these packages installed before installing the library.
sudo apt-get install libgirepository1.0-dev gcc libcairo2-dev pkg-config gir1.2-gtk-3.0 graphviz
brew install gobject-introspection graphviz cairo pkg-config gtk+3 gcc
After installing the packages you can install the library using pypi.
pip install darkhex
Installing from source would be helpful if you wanted a custom version of the library / helping with development. Make sure you have the packages above installed.
We recommend using a virtual environment for the development.
E.g. for Ubuntu or Debian:
sudo apt-get install virtualenv python3-virtualenv
virtualenv venv
source venv/bin/activate
Rest of the installation is just installing python requirements.
E.g. for Ubuntu or Debian:
# Install the libraries listed above
sudo apt-get install libgirepository1.0-dev gcc libcairo2-dev pkg-config gir1.2-gtk-3.0 graphviz
# Install the python packages
pip install -r requirements.txt
Update PYTHONPATH on your .bashrc or .zshrc file (or .venv/bin/activate)