THIS IS WORK IN PROGRESS !! - feel free to comment and improve. some stuff is redundant or optional, some is not (yet) fully implemented..
- my Convolutional-Neural-Network "Rudolf" is learning to play Breakout
- some code was taken from here: https://github.com/Kuax-Meat/Snake-Reinforcement-Deep-Q-Learning
using:
- Tensorflow 1.9.0
- pygame
- python36
- RAMDISK is used for saving NetStates
- plus a Textfile is saved there, for plotting with matplotlib (CNNDiagramm.py)
P.S. I know this code is a mess. sry for that. Im just interested in it. If someone has an idea, your welcome !!
- start with: "python breakout07_train.py" and work through the errors until it is running. ; ) have fun.
- e.g. I have set up a dual graphics-card system. u can get rid of those lines if u do not use two cards.
- ...........Oh my gosh!, it´s actually learning..