Comment by _0ffh

Haven't read the paper yet, but sounds similar in concept to what Geoff Hinton aims at for image recognition networks.