Algorithm in MI-FGSM

[yipeige]

I am not sure if the code for MI-FGSM is correct, since the gradient is expected to be calculated in each iteration rather than a single value for all iterations?

[as791]

Hi @Yip-Ge, yes you are right I think 🤔 . I will try to fix this ASAP.