pralab/Fast-Minimum-Norm-FMN-Attack

Foolbox implementation for NeurIPS 2021 Paper: "Fast Minimum-norm Adversarial Attacks through Adaptive Norm Constraints".

Jupyter NotebookMIT

Issues

why not sign function for gradient under linf norm
#3 opened 4 years ago by Equationliu
2
Loss is summed over the batch to estimate the distance to the boundary
#1 opened 4 years ago by jeromerony
2