AlphaGo-like Machine Learning Model for HOIGI game

In this project, by following the AlphaGo model, we aimed to design a game called Hoigi. We collected over 30000 game positions from games played by different depth levels of the minimax alpha-beta pruning algorithm that we implemented. The data generated that was trained in a supervised learning neural network to predict the favorableness of a position achieved a result of over 94% for test accuracy. To aid the visual, we also self-designed board and pieces and created Portable Game Notation parser to construct game interface.

STACKPython, Pygame, Sklearn, Keras, TensorFlow, Pixel Art.

PlatformGame Interface

GITHUBhttps://github.com/vynpt/AlphaZero-for-Hoigi