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
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 changed | Where | Measured | Effect |
|---|---|---|---|
| Add Jacobian adversarial regularization | CNN · MNIST, FGSM accuracy | accuracy (%): 76.5 -> 98.4 | Improved |
| Add Jacobian adversarial regularization | CNN · MNIST, PGD40 accuracy | accuracy (%): 0 -> 98.1 | Improved |
| Add uniform-noise pixel augmentation | MNIST CNN · MNIST, PGD40 accuracy | accuracy (%): 0 -> 0.02 | No clear effect |
| Add Jacobian adversarial regularization | Wide-ResNet · SVHN, PGD20 accuracy | accuracy (%): 1.96 -> 26.79 | Improved |
| Add uniform-noise pixel augmentation | SVHN Wide-ResNet · SVHN, PGD20 accuracy | accuracy (%): 1.96 -> 9.21 | Improved |
| Add Jacobian adversarial regularization | Wide-ResNet · CIFAR-10, PGD20 accuracy | accuracy (%): 0 -> 15.5 | Improved |
| Add Jacobian adversarial regularization | CIFAR-10 Wide-ResNet · CIFAR-10, PGD20 accuracy | accuracy (%): 7.48 -> 15.5 | Improved |
| Add Jacobian adversarial regularization to one-step adversarial training | FGSM-AT1 Wide-ResNet · CIFAR-10, PGD20 accuracy | accuracy (%): 0.01 -> 46.7 | Improved |
| Combine JARN with one-step adversarial training instead of seven-step PGD adversarial training | CIFAR-10 Wide-ResNet · CIFAR-10, PGD20 accuracy | accuracy (%): 45.9 -> 46.7 | Improved |
| Add JARN to one-step adversarial training | CIFAR-10 Wide-ResNet · CIFAR-10, FGSM accuracy | accuracy (%): 94.5 -> 65.7 | Hurt |
| Use JARN rather than double-backpropagation regularization | CIFAR-10 Wide-ResNet · CIFAR-10, PGD20 accuracy | accuracy (%): 0 -> 15.5 | Improved |
| Use JARN alone rather than PGD-AT7 | CIFAR-10 Wide-ResNet · CIFAR-10, PGD20 accuracy | accuracy (%): 45.9 -> 15.5 | Hurt |
Closest papers by meaning
- Improving DNN Robustness to Adversarial Attacks using Jacobian Regularization (ECCV 2018 · similarity 0.91)
- On Connections Between Regularizations for Improving DNN Robustness (TPAMI 2020 · similarity 0.87)
- Towards the first adversarially robust neural network model on MNIST (ICLR 2019 · similarity 0.87)
- Towards Understanding the Regularization of Adversarial Robustness on Neural Networks (ICML 2020 · similarity 0.87)
- Adversarial Training Can Provably Improve Robustness: Theoretical Analysis of Feature Learning Process Under Structured Data (ICLR 2025 · similarity 0.87)
- Characterizing Model Robustness via Natural Input Gradients (ECCV 2024 · similarity 0.87)
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.
