Minesweeper AI

Various different TensorFlow models and algorithms that play minesweeper.

About

• Model 1
  • An implementation of reinforcement learning that takes the board as an input and attempts to predict the next best move on the board. This model and algorithm was not optimal.
• Model 2
  • An implementation of binary classification that attempts to classify a tile as a mine or not based on the surrounding tiles. This model was significantly more successful than the first.

Run

• Install the required modules: pip install pygame numpy tensorflow
• Use Python 3 to run main.py. This will open a window and have the pre-trained AI play the game. Additionally, you can run the model files directly to train the corresponding model.