Re:A New Lens on Understanding Generalization in Deep Learning
by ersiees on Sunday March 21, 2021 @03:29AM
(#26528738)
The real comparison would have been between Cifar-10 (or Imagenet) and a downsampled version there of, not upsampled, and I actually know for sure that this harms performance.. So this ideal world and real world training are definitely not the same!
[ Parent
| on HN ]
Re:A New Lens on Understanding Generalization in Deep Learning
by preetum on Sunday March 21, 2021 @10:36AM
(#26530756)
This is the exact comparison we make in the paper.
We have subsampled ImageNet experiments as well; see Figure 3 in the full paper: https://arxiv.org/abs/2010.08127