Test results under different standard step size.

<div><p>Adversarial training has become a primary method for enhancing the robustness of deep learning models. In recent years, fast adversarial training methods have gained widespread attention due to their lower computational cost. However, since fast adversarial training uses single-s...

Full description

Saved in:

Bibliographic Details
Main Author:	Jie-Chao Zhao (20521823) (author)
Other Authors:	Jin Ding (11522) (author), Yong-Zhi Sun (20521826) (author), Ping Tan (1356951) (author), Ji-En Ma (20521829) (author), You-Tong Fang (20521832) (author)
Published:	2025
Subjects:	Genetics Molecular Biology Neuroscience Pharmacology Biotechnology Biological Sciences not elsewhere classified Information Systems not elsewhere classified vgg19 models using tiny imagenet datasets making models prone lower computational cost experimental results demonstrate excessive total perturbation deep learning models low accuracy caused prevent catastrophic overfitting avoiding catastrophic overfitting poor robustness due fast adversarial training perturbation step size input clean samples generated adversarial examples catastrophic overfitting clean accuracy adversarial training adversarial examples step size generated based step attacks input noise recent years random noise paper proposes insufficient strength first added existing methods determined based certain shortcomings atss ).
Tags:	Add Tag No Tags, Be the first to tag this record!

Be the first to leave a comment!

Test results under different standard step size.

Similar Items