Perfect work! The implementation of Q-learning is almost perfect and the results are obvious. The code is also clear and the use of player interfaces means that it is easy to maintain and extend.
Improvements
A possible improvement could be to vary the learning rate and discount factor during training rather than keeping them constant.
Apart from this small detail everything seems to me to be perfect and working, and the results are no less.
Best regards,
Alessandro
Overall Feedback
Perfect work! The implementation of Q-learning is almost perfect and the results are obvious. The code is also clear and the use of player interfaces means that it is easy to maintain and extend.
Improvements
A possible improvement could be to vary the learning rate and discount factor during training rather than keeping them constant.
Apart from this small detail everything seems to me to be perfect and working, and the results are no less. Best regards, Alessandro