You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
If you read the paper, you'll find that the version of BPDA they attempted is not the generalized version. It does not implement the generalized BPDA that attacks individual layers. Nor does the original paper describing BPDA go into detail on how to discover the approximation function g.
BPDA itself isn't an attack, but a helper for other attacks that requires gradients - BPDA is way to approximate those gradients when the function is not differentiable.
Hi there,
Could you please let me know where can I find this function? It doesn't look like it is included in the current cleverhans package. Thanks!
The text was updated successfully, but these errors were encountered: