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
Hi I am relatively new to coding.
I recently switched from Foolbox 2.4.0 to the newer version 3.0.4.
In the previous version I was mostly using the FGSM or the PGD attack without the need to specify the epsilon value.
I noticed that in the new version I need to specify the espilon value/s within the attack()
Is there any way to avoid setting the epsilon value/s beforehand, and letting the attack compute a specific(possibly different) minimum epsilon value for each distinct pixel?
Thanks
The text was updated successfully, but these errors were encountered:
it depends whether the attack you're trying to run is a minimization attack (CW for example) or a fixed epsilon (LinfPGD) attack. If it is a minimization attack by default, then you can just set the epsilons to None. Otherwise you have to specify the epsilon value. In this case, I suggest you try running the attack with a wide range of epsilons in order to find a suitable one.
Hi I am relatively new to coding.
I recently switched from Foolbox 2.4.0 to the newer version 3.0.4.
In the previous version I was mostly using the FGSM or the PGD attack without the need to specify the epsilon value.
I noticed that in the new version I need to specify the espilon value/s within the attack()
Is there any way to avoid setting the epsilon value/s beforehand, and letting the attack compute a specific(possibly different) minimum epsilon value for each distinct pixel?
Thanks
The text was updated successfully, but these errors were encountered: