First of all, you have done a nice job to complete the lab with very nice details. Quite clever to spot that the optimal strategy provided is indeed not optimal and replace it with an implementation which i believe is the true strategy to win the Nim game. Further than that, the convergence of weights towards the optimal strategy has proved that the evolution process works well. However, i think there is still much room for improvements. The optimal strategy can be excluded and replaced by a set of more basic rules to test the ability of ES algorithms to find a good strategy as a combination of basic rules in case an optimal one is absent. Good job anyway. Hope to collaborate with you in the future.
First of all, you have done a nice job to complete the lab with very nice details. Quite clever to spot that the optimal strategy provided is indeed not optimal and replace it with an implementation which i believe is the true strategy to win the Nim game. Further than that, the convergence of weights towards the optimal strategy has proved that the evolution process works well. However, i think there is still much room for improvements. The optimal strategy can be excluded and replaced by a set of more basic rules to test the ability of ES algorithms to find a good strategy as a combination of basic rules in case an optimal one is absent. Good job anyway. Hope to collaborate with you in the future.