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
Dear Minhao Cheng:
When I check the code, I found the K (which is used for estimating the gradient of θ), is set different in Sign-OPT and SVM-OPT?
100 vs 200
I think this is not fair for comparing Sign-OPT and SVM-OPT
This is the code in def sign_grad_v1, 200 is the default value.
K = 100 as the default parameter of the def sign_grad_svm function, which does not pass in any parameter?
Should I set all K to100 and redo my experiments for fair comparison of Sign-OPT and SVM-OPT in my paper?
Thank you!
The text was updated successfully, but these errors were encountered:
Dear Minhao Cheng:
When I check the code, I found the K (which is used for estimating the gradient of
θ
), is set different in Sign-OPT and SVM-OPT?100 vs 200
I think this is not fair for comparing Sign-OPT and SVM-OPT
This is the code in
def sign_grad_v1
, 200 is the default value.attackbox/attack/Sign_OPT.py
Line 201 in 1f82092
K = self.k # 200 random directions (for estimating the gradient)
and in
attackbox/attack/Sign_OPT.py
Line 277 in 1f82092
K = 100
as the default parameter of thedef sign_grad_svm
function, which does not pass in any parameter?Should I set all K to100 and redo my experiments for fair comparison of Sign-OPT and SVM-OPT in my paper?
Thank you!
The text was updated successfully, but these errors were encountered: