Christoph Franke
Sep 7, 2018 · 1 min read

Thank you for this very interesting article. I am trying to reproduce your results and I find that after some iterations the AI is able to win the game once. However, if you then load the succcess.model from disk and let it replay it will not be able to win the game again, presumably because the agent is nondeterministic due to the exploring epsilon. Have you investigated further and come up with a solution? If so, I’d be very interested to hear!

    Christoph Franke

    Written by