I finally got off my duff and wrote up a graphical version of the game, and promptly trashed the AI that had previously beaten me. It's amazing what you can do when you're paying attention! As a side benefit to doing the graphic version, it's now much easier to share the results of games. OK, … Continue reading Much better
Well, that was embarrassing. I hadn't played the AI for a while, as I was waiting until I'd written a better version for display. Anyway, I tried a game against the 5x5000* and, due to not paying attention, I lost. He lulled me into a false sense of security, making several repeated plays on vertical … Continue reading Sarah Conner?…
So I tried an experiment: For a few rounds, I generated my training data by only picking the best performing neural network, playing it against the Smart Random player, and only adding those games where the NN lost to the data set. It was, quite bluntly, an unmitigated disaster. Every NN began performing worse by … Continue reading Training data matters… and it’s hard to make
After writing yesterday, I ran a few rounds through the evening and overnight (god bless NVidia. My GTX1030 finished four rounds of training in that time frame). Unfortunately, I can now confirm that I've reached the limit of what I can train with my current method of generating training data. Some background: Neural Networks, as … Continue reading Getting better training data
Skynetting Myself 