Dah, that's nothing. We have models that we built specifically to determine the best way to train other models. After 2 years of tweaking these models, we learned that the best way to optimize our learning was to build new models that train these models for determining how to train our other models.