Material from the "Lets design neural networks for playing" pong workshop offered at the "Forschungstage Informatik 2023" in Saarbrücken.
The workshop consists of two parts:
- Understanding how neural networks structures change their learning behaviour (using the tensorflow playground).
- Implementing core components required for training pong with a deep reinforcement learning algorithm.
Note that the workshop as given contains multiple discussion rounds, for which there are no guidelines contained in this repository.
However, if you want to use this material, note that almost certainly the contents of this repository is too extensive and must be reduced.
You can find possible exercises to complement the jupyter notebook here.
This repository will be soon updated to be more guided; also the jupyter notebook will be split into multiple individual notebooks.