adaptive_attacks_paper icon indicating copy to clipboard operation
adaptive_attacks_paper copied to clipboard

Published 20 hours ago • verified publisher icon wielandbrendel

Implementation of the remaining experiments from the paper

Open mipet23 opened this issue 4 years ago • 5 comments

Hi, thanks for providing the code. Where can I find the other implementations of the remaining experiments of your paper?

mipet23 avatar Dec 25 '20 16:12 mipet23

@carlini will add his implementations soon, then we should be complete.

wielandbrendel avatar Dec 31 '20 15:12 wielandbrendel

Hi, thanks for your excellent work, but could you complete this repo soon? @wielandbrendel @carlini

jiangyangzhou avatar Mar 08 '21 08:03 jiangyangzhou

I really appreciate your excellent paper and work. @wielandbrendel @carlini I think I can learn more from this code and the other incomplete implementations.
I tried to attack MMC with the regression loss $|| f(x) - mu ||^2_2$, the results are comparable to the results in the MMC paper. So I'm really interested in how I can attack the MMC model to ~0.5% robust acc.

sndnyang avatar Apr 16 '21 15:04 sndnyang

I really appreciate your excellent paper and work. @wielandbrendel @carlini I think I can learn more from this code and the other incomplete implementations. I tried to attack MMC with the regression loss $|| f(x) - mu ||^2_2$, the results are comparable to the results in the MMC paper. So I'm really interested in how I can attack the MMC model to ~0.5% robust acc.

Oh, I think I get a similar result, while MMC seems not strong but I could not attack MMC to 0.5%, and far from that.

jiangyangzhou avatar Apr 16 '21 15:04 jiangyangzhou

@carlini Is there still a chance, that the remaining implementations will be published here?

oGkgN avatar Apr 25 '22 10:04 oGkgN