❔ Any questions
In the original paper (https://arxiv.org/abs/2003.01690), the authors mention that they used 100 iterations for each run of the white-box attacks. However, in the current implementation, it appears that the attacks are only run for 10 iterations. Can anyone explain why this change was made? Is there a trade-off between accuracy and efficiency?
❔ Any questions
In the original paper (https://arxiv.org/abs/2003.01690), the authors mention that they used 100 iterations for each run of the white-box attacks. However, in the current implementation, it appears that the attacks are only run for 10 iterations. Can anyone explain why this change was made? Is there a trade-off between accuracy and efficiency?