Connect 4 game with AI, written in Python. I tried to make an AI which is unbeatable in "connect-4" with Deep Learning methods.
git clone https://github.com/cansozbir/Connect-4.git /type/your/path
pip install -r requirements.txt
python3 /type/your/path/main.py
Model's architecture : input > FC(42) > BatchNorm > FC(64) > FC(64) > FC(64) > FC(42) > FC(3,softmax) > output
Model's validation accuracy is %99 but it doesn't play the game good enough. And I think that's because of the dataset. It's created with ~9 million random games but i'm not sure about it's homogeneity. If you have any suggestions, let me know! Dataset : https://drive.google.com/open?id=17XbCd5BL20AJLTb_oAuKcZrhyHHHeMli
