AI Paper Atlas

Research map / Adversarial robustness and security / Certified and empirical robustness

Jacobian Adversarially Regularized Networks for Robustness

Alvin Chan, Yi Tay, Yew Soon Ong, Jie Fu

ICLR 2020 · Adversarial robustness and security › Certified and empirical robustness · approach: Regularized defenses against attacks · 8 citations (OpenAlex)

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
Add Jacobian adversarial regularizationCNN · MNIST, FGSM accuracyaccuracy (%): 76.5 -> 98.4Improved
Add Jacobian adversarial regularizationCNN · MNIST, PGD40 accuracyaccuracy (%): 0 -> 98.1Improved
Add uniform-noise pixel augmentationMNIST CNN · MNIST, PGD40 accuracyaccuracy (%): 0 -> 0.02No clear effect
Add Jacobian adversarial regularizationWide-ResNet · SVHN, PGD20 accuracyaccuracy (%): 1.96 -> 26.79Improved
Add uniform-noise pixel augmentationSVHN Wide-ResNet · SVHN, PGD20 accuracyaccuracy (%): 1.96 -> 9.21Improved
Add Jacobian adversarial regularizationWide-ResNet · CIFAR-10, PGD20 accuracyaccuracy (%): 0 -> 15.5Improved
Add Jacobian adversarial regularizationCIFAR-10 Wide-ResNet · CIFAR-10, PGD20 accuracyaccuracy (%): 7.48 -> 15.5Improved
Add Jacobian adversarial regularization to one-step adversarial trainingFGSM-AT1 Wide-ResNet · CIFAR-10, PGD20 accuracyaccuracy (%): 0.01 -> 46.7Improved
Combine JARN with one-step adversarial training instead of seven-step PGD adversarial trainingCIFAR-10 Wide-ResNet · CIFAR-10, PGD20 accuracyaccuracy (%): 45.9 -> 46.7Improved
Add JARN to one-step adversarial trainingCIFAR-10 Wide-ResNet · CIFAR-10, FGSM accuracyaccuracy (%): 94.5 -> 65.7Hurt
Use JARN rather than double-backpropagation regularizationCIFAR-10 Wide-ResNet · CIFAR-10, PGD20 accuracyaccuracy (%): 0 -> 15.5Improved
Use JARN alone rather than PGD-AT7CIFAR-10 Wide-ResNet · CIFAR-10, PGD20 accuracyaccuracy (%): 45.9 -> 15.5Hurt

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy