Research map / Adversarial robustness and security / Certified and empirical robustness
Adversarial Learning With Margin-Based Triplet Embedding Regularization
Yaoyao Zhong, Weihong Deng
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 MTER regularization | Softmax ResNet-18 · MNIST BIM | accuracy under attack (%): 0.0 -> 98.7 | Improved |
| add MTER regularization | ArcFace ResNet-18 · MNIST ITGSM | accuracy under attack (%): 24.7 -> 95.3 | Improved |
| replace adversarial training with MTER | Softmax ResNet-18 · MNIST BIM | accuracy under attack (%): 3.1 -> 98.7 | Improved |
| replace adversarial training with MTER | Softmax ResNet-18 · MNIST FGSM | accuracy under attack (%): 99.9 -> 96.8 | Hurt |
| add MTER regularization | Softmax ResNet-18 · MNIST clean | accuracy (%): 99.6 -> 99.5 | No clear effect |
| add MTER finetuning | ArcFace ResNet-50 · Face disguise IFTGSM eps=10 (CASIA-WebFace) | hit rate (%): 100.0 -> 27.4 | Improved |
| add MTER finetuning | ArcFace ResNet-50 · Face disguise IFTGSM eps=10 (MS1M) | hit rate (%): 100.0 -> 61.4 | Improved |
| finetune on MS1M+VGGFace2 instead of MS1M only with MTER | ArcFace ResNet-50 · Face disguise IFTGSM eps=10 | hit rate (%): 61.4 -> 19.5 | Improved |
| add MTER finetuning | ArcFace ResNet-50 · LFW (trained on CASIA-WebFace) | accuracy (%): 99.5 -> 99.5 | No clear effect |
| replace adversarial finetuning with MTER | ArcFace ResNet-50 · Face disguise IFTGSM eps=10 (CASIA-WebFace) | hit rate (%): 49.5 -> 27.4 | Improved |
| replace adversarial finetuning with MTER | ArcFace ResNet-50 · Face disguise IFTGSM eps=10 (MS1M) | hit rate (%): 62.6 -> 61.4 | No clear effect |
Closest papers by meaning
- Adversarial Margin Maximization Networks (TPAMI 2019 · similarity 0.90)
- Towards Robustness of Deep Neural Networks via Regularization (ICCV 2021 · similarity 0.88)
- Improved Sample Complexities for Deep Neural Networks and Robust Classification via an All-Layer Margin (ICLR 2020 · similarity 0.86)
- Boosting Adversarial Training via Fisher-Rao Norm-based Regularization (CVPR 2024 · similarity 0.85)
- Rethinking Invariance Regularization in Adversarial Training to Improve Robustness-Accuracy Trade-off (ICLR 2025 · similarity 0.85)
- Kernel Learning with Adversarial Features: Numerical Efficiency and Adaptive Regularization (NeurIPS 2025 · similarity 0.85)
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.
