A agent for Slither.io,
which is an online game,using actor-critic algorithms.
To conquer the uncontrollable difficulty degree,
dis-converge or fall into local minima, we propose four methods and greatly improve the performance.
For more detail, please see our report:Report
We don't upload our model, so please train before test.
- Train:
python train_AC.py - Test:
python play_AC.py
There might have some bug or unstable since it's our experiment code.
