The idea behind this is to tinker with some Tensorflow, Keras and Pygame to implement a Deep Q Learning model. These models are good for understanding a set of undefined rules based on an action/reward feedback loop based on a current state.
Run the game to see the Marine shoot at the Alien while it learns to doge the bullets more effectivley.
pipenv install
python run app.py --help
Good reference codebase: https://github.com/maurock/snake-ga/