Simple pacman, with machine learning.
NEAT algorithm from https://github.com/xviniette/FlappyLearning
Pacman sprites from https://www.spriters-resource.com/arcade/pacman/sheet/52631/
Pacman sounds from https://github.com/masonicGIT/pacman
cd clientnpm installcd ../servernpm installnpm run dev- go to localhost
Machine learning using neuroevolution to perform pathfinding for pacman.
Currently, it tries to learn using a single ghost, blinky, and the resulting score is the amount of dots eaten.
By using current settings, it is able to get around 120 dots eaten at around generation 50-150.
Given a longer period of time, around 1000 gens, it seems that the network can be improved to around 200 dots.
Control volume, pause and resume, and play speed with dat.gui controller.