Accelerating Adversarial Perturbation by 50% with Semi-backward Propagation

Zhiqi Bu

Accelerating Adversarial Perturbation by 50% with Semi-backward Propagation

Machine Learning 2022-11-10 v1 Optimization and Control

Authors: Zhiqi Bu

Abstract

Adversarial perturbation plays a significant role in the field of adversarial robustness, which solves a maximization problem over the input data. We show that the backward propagation of such optimization can accelerate $2\times$ (and thus the overall optimization including the forward propagation can accelerate $1.5\times$ ), without any utility drop, if we only compute the output gradient but not the parameter gradient during the backward propagation.

Keywords

adversarial training adversarial examples

Cite

@article{arxiv.2211.04973,
  title  = {Accelerating Adversarial Perturbation by 50% with Semi-backward Propagation},
  author = {Zhiqi Bu},
  journal= {arXiv preprint arXiv:2211.04973},
  year   = {2022}
}

Related papers

View all related →