The Snake-AI project combines reinforcement learning with a feedforward network to create a self-improving AI that will learn to play the game Snake. Several different controlled experiments (see comments in entities.py for details) were done to lower the number of iterations necessary to reach a "good" solution. The network's logic, in logic.py, is my own implementation of a feedforward network. It has not been optimized and is used for personal practice.
This project uses pygame and numpy.
Please run pip install pygame
and pip install numpy
to ensure that these are installed.
python game.py