Release Date?
Thanks for your attention. In my research two years ago, a single layer neural networks gun is to it's greatest extent as good as my knn gun (only against WaveSurfers), and with more layers (fully connected) the performance is even worse. But the success of Gaff against WaveSurfers still appeals to me, and I'm trying hard to match his result, whether with the same settings or with different approach, e.g. Convolutional Neural Networks, which is popular nowadays. I think I may not release it until it beats DrussGT ;) Try to change my mind ;)
I would put it in competition. Even if you think it is weak now (which I doubt giving the beauty of the U api), it is fun to see it improving. Also, it gives newcomers the sense that the robocode is still alive.
Me too. Without putting it in the competition you can't know what robots you do bad against so improving would be harder.(1150 bots) Also I want to see another active NN bot=).