apparently adaptive sampling is used in machine learning too i haven't looke this up, but considering it i'm imagining thinking of the layers of a transform as arbitrarily warped n-dimensional spaces that transform some input to some output once you think of them as spaces, adaptive sampling makes more sense, as you coul subivie the spaces based onwhere changes are, and you could kind of imagine making arbitrarily large models where most of the contents are elided into interpolation between known things, but i dunno if that would really work as it seems like the points where things change are the points of interest, dunno, maybe i am wrong but that was the big websearch hit which is why i added path tracing to the query :s i wonder what it means to use it in machine learning