By: Sean Weisner (saw336), Diego Virtue (dtv25), Frank Zhou (fcz5)
Link to paper: https://docs.google.com/document/d/18SRsW-SpH7Gs73xyqAHALdgLe8Q-qOZ2oVjYi7hlL-0/edit?usp=sharing
This README.md serves as a guide for how to run our code.
Please run the following commands:
pip install sympy
pip install numpy
pip install matplotlib
pip install contextlib
pip install unittest
pip install torch
pip install torchvision
To "build" a variation of the probabilistic model, run minesweeper.py and follow the instructions provided by the terminal.
We suggest running the following parameters for board size (n x n) and mine count (m) based on the following difficulties:
Beginner: n = 9, m = 10
Intermediate: n = 16, m = 40
Expert: n = 22, m = 99
Then specify if you want to use the Single-Point or Constraint Satisfaction alogirthm to run, and then specify if you want a random tile to open the tile with the lowest local probability given the need to make an uncertain move.
To train a DQN: run agent.py in the DeepQModel directory. You can select the game board size by modifying the NUM_TILES variable on line 17, and you can select the number of mines by modifying the NUM_MINES variable on line 18. If you would like to change the number of game iterations that the model trains on, edit the num_episodes variable on line 148 if you plan to train on your GPU, and edit the same variable on line 150 if you plan to train on your CPU. If you would like to save the model, change the name of the model on line 196 and run the code. Your model will be saved in the DeepQModel directory.