|
@tanmingxing | |||||
|
Can adversarial examples improve image recognition? Check out our recent work: AdvProp, achieving ImageNet top-1 accuracy 85.5% (no extra data) with adversarial examples!
Arxiv: arxiv.org/abs/1911.09665
Checkpoints: git.io/JeopW pic.twitter.com/bAu054LGt2
|
||||||
|
||||||
|
Kirill Klimov
@klimov_k_v
|
26. stu |
|
Awesome paper! May i ask do auxiliary BNs shares learnable parameters? Or BNs have not only separate moving average statistics but separate learnable scales/offsets too?
|
||
|
||
|
|
Mingxing Tan
@tanmingxing
|
26. stu |
|
Hi Kirill, good question! we use separate scale/offset and stats.
|
||
|
|
||
|
Rohan Taori
@rtaori13
|
25. stu |
|
Do you also evaluate the adversarial accuracy of the models? How does it compare to the baseline?
|
||
|
|
||