minesweeper-ai

Here I am trying to train AI to play Minesweeper.

I also included a working implementation of Minesweeper game using Tkinter GUI package.

To play the game, run python src/game.py

The Player AI is implemented in the src/Minesweeper_Conv.ipynb notebook.

Visualization

White cells - high confidence that the cell has no mine.
Yellow - no confidence.
Dark Red - high probability of a mine.
Black - high confidence that it has a mine or cell that is already open.

Files:

• engine.py - The business logic of the game.
• game.py - GUI for playing it (requires tkinter package)
• ml_dataset.py - Helper functions and classes for generating training datasets for our ML model.
• Minesweeper_Conv.ipynb - Notebook with the model definition and the training loop.

Dataset

Each of our training samples is a partially opened game board where numbers 0-8 are stored in each open position and 10 indicates hidden positions.

For example:

[[10, 10, 10, 10, 10, 10, 10, 10, 10, 10],
 [10, 10, 10, 10, 10, 10, 10, 10, 10, 10],
 [ 1, 10, 10,  5, 10, 10, 10, 10, 10, 10],
 [10, 10, 10, 10, 10, 10, 10, 10, 10, 10],
 [10, 10, 10, 10,  4, 10, 10, 10, 10, 10],
 [10, 10, 10, 10, 10, 10, 10,  3,  1,  1],
 [10, 10, 10, 10, 10, 10, 10,  2,  0,  0],
 [10, 10, 10, 10, 10, 10, 10,  2,  0,  0],
 [10, 10, 10, 10,  3,  3,  3,  3,  2,  1],
 [10, 10, 10, 10,  1,  0,  1, 10, 10, 10],
 [10, 10, 10, 10,  2,  2,  3, 10, 10, 10],
 [10, 10,  1, 10, 10, 10, 10, 10, 10, 10]]

is a tensor representation of the following board:

☐ ☐ ☐ ☐ ☐ ☐ ☐ ☐ ☐ ☐
1 ☐ ☐ 5 ☐ ☐ ☐ ☐ ☐ ☐
☐ ☐ ☐ ☐ ☐ ☐ ☐ ☐ ☐ ☐
☐ ☐ ☐ ☐ 4 ☐ ☐ ☐ ☐ ☐
☐ ☐ ☐ ☐ ☐ ☐ ☐ 3 1 1
☐ ☐ ☐ ☐ ☐ ☐ ☐ 2    
☐ ☐ ☐ ☐ ☐ ☐ ☐ 2    
☐ ☐ ☐ ☐ 3 3 3 3 2 1
☐ ☐ ☐ ☐ 1   1 ☐ ☐ ☐
☐ ☐ ☐ ☐ 2 2 3 ☐ ☐ ☐
☐ ☐ 1 ☐ ☐ ☐ ☐ ☐ ☐ ☐

Labels are stored as a matrix where 1 is a position with a mine, 0 is a position without mine, and -100 if the cell is open (cell_value != 10 in the input tensor). Note that -100 is a special value for F.cross_entropy() that tells it not to compute loss in that cell (because it's open already and its value is known via the input).

For example:

[[   0,    0,    0,    1,    0,    0,    0,    1,    1,    0],
 [   0,    0,    0,    1,    0,    0,    0,    1,    0,    1],
 [-100,    0,    0, -100,    1,    0,    0,    0,    0,    1],
 [   1,    0,    1,    1,    1,    1,    1,    1,    1,    1],
 [   0,    1,    0,    1, -100,    0,    1,    0,    1,    0],
 [   0,    1,    0,    0,    0,    0,    0, -100, -100, -100],
 [   1,    1,    0,    0,    1,    0,    1, -100, -100, -100],
 [   0,    0,    1,    0,    1,    1,    1, -100, -100, -100],
 [   0,    0,    1,    0, -100, -100, -100, -100, -100, -100],
 [   0,    1,    0,    1, -100, -100, -100,    1,    1,    0],
 [   1,    0,    1,    0, -100, -100, -100,    0,    0,    0],
 [   0,    0, -100,    0,    0,    1,    1,    0,    1,    0]]