Hi Federico here is my review for your work, I hope you will appreciate it.
First of all, your work seems well done to me and your extension of what we saw in class using Q-learning seems correct.
I have some pointers for you to improve your work and make it better understandable.
I would suggest you to divide your code into classes by perhaps creating a "game" class and a "player" class this allows for better organization and also better reading of the code.
In the training phase you could add an additional "epsilon" parameter to encourage agent exploration by choosing a random action (greedy approach)
The results look good to me, try to increase the number of matches to update the Q-Table you might have better results.
To make the agent more robust, you could do training with different types of players, to be varied randomly during the various iterations just for this reason, creating a player class to be extended could be a good idea.
In conclusion your work seems well done, I only suggest you to improve a bit the organization of your code.
Good work for the next projects!
Hi Federico here is my review for your work, I hope you will appreciate it.
First of all, your work seems well done to me and your extension of what we saw in class using Q-learning seems correct. I have some pointers for you to improve your work and make it better understandable. I would suggest you to divide your code into classes by perhaps creating a "game" class and a "player" class this allows for better organization and also better reading of the code. In the training phase you could add an additional "epsilon" parameter to encourage agent exploration by choosing a random action (greedy approach) The results look good to me, try to increase the number of matches to update the Q-Table you might have better results. To make the agent more robust, you could do training with different types of players, to be varied randomly during the various iterations just for this reason, creating a player class to be extended could be a good idea.
In conclusion your work seems well done, I only suggest you to improve a bit the organization of your code. Good work for the next projects!