… leveraging the regularization power of adversarialexamples. … wrt the network parameter for gradient updates. In other words, … The family of EfficientNets provides a strong baseline, eg., …
C Finlay, AM Oberman - Machine Learning with Applications, 2021 - Elsevier
… that gradientregularization does not lead to gradient obfuscation or gradient masking. … or minimized, so that gradient-based attacks fail to produce adversarialexamples. For …
… Compared with previous widely used model-dependent methods such as fast gradient sign (FGS) [9] and the state-ofthe-art method CW [20], which utilize a well-trained classifier to …
… Such modified inputs, also known as adversarialexamples (AEs), pose a doubt when applying deep learning models to security-sensitive applications, such as face recognition, …
… projected gradient ascent based adversarial … norm regularization. This fundamental connection confirms the long-standing argument that a network’s sensitivity to adversarialexamples …
… adversarial training with multi-task cues. In our experiments, we find that the deep gradient regularization of … While there are a variety of ways to generate adversarialsamples, the fastest …
CH Ho, N Nvasconcelos - Advances in Neural Information …, 2020 - proceedings.neurips.cc
… a new family of adversarialexamples for constrastive … attacks with the popular fast gradient sign method (FGSM) [19]. … Virtual adversarial training: A regularization method for supervised …
Y Wen, S Li, K Jia - International Conference on Machine …, 2020 - proceedings.mlr.press
… The problem of adversarialexamples has shown that modern Neural Network (NN) models could be rather fragile. Among the more established techniques to solve the problem, one is …
… , we develop a novel family of algorithms that generalize the AT … benign examples when crafting adversarialexamples due to the … Virtual adversarial training: a regularization method for …