Skip to content

Latest commit

 

History

History
11 lines (8 loc) · 951 Bytes

README.md

File metadata and controls

11 lines (8 loc) · 951 Bytes

AI Game Routing

A more in-depth fork of https://github.com/minerharry/neat_game_testing, attempting to teach a neural network how to play Super Mario Bros 1.

NOTE: for the game, I am using marblexu's (https://github.com/marblexu) awesome Python Mario Bros game, at https://github.com/marblexu/PythonSuperMario/. I have torn it apart to optimize and fit it with my infrastructure :).

Streamed on https://www.twitch.tv/minerharry 24/7 when running. Some extra details about the training progress can be found in the panels below the stream.

Resources: https://neat-python.readthedocs.io/en/latest/neat_overview.html - the python library I am using to run these tests http://www.cs.ucf.edu/~kstanley/#publications - papers about neat, including the original that started the whole thing http://nn.cs.utexas.edu/downloads/papers/stanley.ec02.pdf - a specific paper in the above list with some helpful information about the reproduction algorithm