devastating new research showing serious problems with robustness in deep nets by @beenwrekt; fits perfectly with the recent @bethgelab work, and my own critical appraisal of deep learning.https://twitter.com/maxalittle/status/1003650956716855296 …
-
You are hyping anti-hype. It's a really cool paper, but calling it 'devastating' isn't particularly helpful. For example, ResNets (which have been the standard for a few years!) seem a lot more robust.
-
-
I should have worded that more carefully: they still suffer from generalization loss, but they perform quite well even on the new dataset. They are still overfit to the original dataset.
- 1 more reply
New conversation -
Loading seems to be taking a while.
Twitter may be over capacity or experiencing a momentary hiccup. Try again or visit Twitter Status for more information.