Minesweeper + Rust + Machine learning ?

This is a rather simple machinelearning project, made a lot harder by writing everything in rust. Yep, even the machine learning part.

Project Background

I can't lie, im a rust fanboy. So ever since i stumbled upon burn-rs i've wanted to try it out. Now, minesweeper is a rather weird choice, one could of course just write some simple rules that would solve most minesweeper board, but this is more about exploring burn than solving a hard problem.

Limitations

• Burn features hotswappable backends, but currently this project is only using Cuda.
• Its not very pretty, doesnt work very good either. (Oh well, it does predict somewhat good)

How to run

1. Clone the repo

git clone [email protected]:thomsen85/rusty-minesweeper.git

2. Train your model

cargo run --bin cuda-train --release

3. Start the app

cargo run --bin minesweeper