We consider a rectangular (NY-like) city of 50×50 crossings.There are thus 2500 states and 4 actions.
- python > 3.6
- numpy
- numba
- matplotlib
- seaborn
python model-based.py && python q_learning.py
Q-learning is written entirely in numpy so it can be optimized using numba library. Using numba gives ~9times learning process speedup.