How many people think the recent ForestNet paper is all gas? https://arxiv.org/pdf/1702.08835.pdf … 1/npic.twitter.com/0EqR3cH8m1
-
7/n Talking of small datasets, who does an MNIST eval these days? And if you do, is LeNet '95 the best you got?pic.twitter.com/tmDATfyBbI
it's a disingenuous paper. You don't compare a large ensemble of models to a single classifier from the 90s. On MNIST!
