Neural Networks playing snake game trained by genetic algorithm
A personal project made by Robin Mancini and myself, consisting in training neural networks to play the game "Snake"
This repository contains:
- The Snake Game itself
- A Genetic Algorithm module
- A Neural Network module
- A main file with toy examples
I timed most functions to be sure to improve speed and used numba jit for compiling some functions, the genetic algorithm is parallelized for its main part (snakes evaluation) using multiprocessing and joblib
Python 3 was used for this project and I can't promess that older versions are compatibles
Libraries you'll need to run the project:
{joblib, numpy, numba, pygame}
Clone the repo using
git clone https://github.com/valentinmace/snake.gitYou will find some ready to run examples in main.py file.
You can try to:
- Play snake
- Train your own neural networks (it can take a while to get good results)
- Display a game played by neural networks that I trained and selected four you
Everything is explaind in the file, just uncomment parts that you want to execute, then go to terminal and do:
python main.pyDo not hesitate to contact me if you need some help
Everything is made by me, I did not want to use existing framework for the genetic algorithm or neural network for learning purposes. I also coded the game with performance in mind rather than conception elegance.
I have published (or will publish depending on when you read this) a serie of youtube tutorial videos on my channel (in french)
Valentin Macé – LinkedIn – YouTube – Twitter - valentin.mace@kedgebs.com
Distributed under the MIT license. See LICENSE for more information.