Last week we looked at the information bottleneck theory of DNNs. This ICLR'18 submission provides a critical analysis - turns out the choice of activation function has/had a lot to do with it!http://blog.acolyer.org/2017/11/24/on-the-information-bottleneck-theory-of-deep-learning …
-
-
Also unlike with preprints, which have names and DOIs or some ID (arXiv), the public PDF on this setup has neither an ID nor author names. It's open to abuse, if understand it correctly since I can see malevolent actors as well as just mistakes leading to copying anonymous' work.
-
I suppose it's not being abused since it's still in use. But open at the time of commenting and double blinding seems a strange combination to me, for the reasons I just mentioned.
-
I also noticed they allow preprints. What's the point of double blind if they allow preprints? If the point is to make it a level playing field then obviously a lab with a good reputation will "unlevel" the playing field by unanonymising themselves with a preprint.
-
Anyway, it's just my thoughts on this since I've never seen this combination of open at the time of review and double blind. Thanks for explaining it. If I ever submit to such a conference I'll be prepared!

-
Obviously it’s unblinded if accepted. So it’s just this weird period right now of being able to talk about it, but not cite it. In normal circumstances I would wait to cover such a paper, this one is an odd exception.
-
Yeah, if not accepted though it's in a kind of limbo. And even more vulnerable, I'd say.
-
That’s a really good point. Zombie papers!
End of conversation
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.