The Froggerithm

As machine learning has become more and more ubiquitous over recent years, it has come to play a greater and greater role in our lives. From the social media posts we see, to the systems that monitor bank accounts for financial fraud, this technology plays an important role in many of the technological systems of the day. This project has no such high-minded goals, however. Our algorithm will learn to do one thing, and one thing only.

Play Frogger.

This project will seek to recreate the classic arcade game Frogger, and then build a machine learning AI to play it through the use of a genetic training algorithm and neural network. The game and the AI will be built in Python, using Pygame and the NEAT-python library.

Prerequisites

Python 3.7 (Note: newer versions will not work)

Initial Setup

Clone repository
Open terminal on repository root directory
python --version to verify python 3.7
python -m venv venv && "venv\Scripts\activate.bat" && pip install -r requirements.txt to install dependencies
python src\main.py -train <generations> to train the neural network
<generations> determines the number of generations the neural network will be trained for using the NEAT algorithm

AI Training

This project makes use of the NEAT-python library to handle most of the heavy lifting for the neural network. The documentation for the NEAT-python library can be found here. The NEAT configuration settings are stored in src/neat_config.txt. Descriptions of the config parameters and what they do can be found in the NEAT-python documentation. Population checkpoints are saved every 100 generations and every 20 minutes in src/Checkpoints. If you wish to load a population checkpoint rather than train from scratch, you can do so by uncommenting the line in run_neat() inside the main method in main.py, which is labelled with a comment.

Documentation

This project includes pydocs for all major functions and classes. The pdoc3 Python tool was used to auto-generate an HTML tree containing all of the pydocs in the project which can be viewed in any major web browser by loading doc/src/index.html. Any time a pydoc is changed in the project, ensure that this tree is re-generated using pydoc3. There is no need to rebuild for folders that do not contain changes to the pydoc.

Unit Testing

This project will primarily make use of the built-in Python unit testing framework. All unit testing source should be placed in src\Unittests.

Extra Credits

Logo created by Azariah Emond