I looked at how many games total each network won during a given round, and came up with the following: The first thing to notice is that there is a LOT of variability here. Individual standings are all over the map. That being said, the top standings seem to be dominated by the deeper networks. … Continue reading First Five, more analysis
Yesterday I ran a few more rounds of training. Here are the results: One thing I've been wondering about is, because I am able to complete a training session so quickly, whether I should either increase the size of the training data set, or increase the number of epochs I train for. For now, I'm … Continue reading Rounds 3,4, and 5 Raw results
Results for round 2 are in and, while there aren't any major changes (results are still fairly random, each AI winning between 1/3 to 2/3 of their matches), what did surprise me was that #3, 1x1000, actually beat the smart random player. He had first move advantage, but it's still an impressive stroke of luck. … Continue reading Round 2 Results
So that's a surprise... I estimated that training would take eight hours, and it took 3 1/2. Just goes to show, you have to be careful extrapolating from small data sets. So I started over and ran the first round again. Here are the results: None of the AIs won fewer than 20 games, or … Continue reading Round 1 Redux
Skynetting Myself 