Research map / Adversarial robustness and security / Certified and empirical robustness
Generalization Bounds for Adversarial Contrastive Learning
Xin Zou, Weiwei Liu
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 Frobenius-norm regularizer lambda=0.002 (ablation) | adversarial contrastive learning (AERM) · CIFAR-10, PGD eps=0.01 | adversarial accuracy: 67.67 -> 69.25 | Improved |
| add Frobenius-norm regularizer lambda=0.002 (ablation) | adversarial contrastive learning (AERM) · CIFAR-10, PGD eps=0.02 | adversarial accuracy: 46.71 -> 48.17 | Improved |
| add Frobenius-norm regularizer lambda=0.05 (ablation) | adversarial contrastive learning (AERM) · CIFAR-10, FGSM eps=0.01 | adversarial accuracy: 67.28 -> 68.8 | Improved |
| add Frobenius-norm regularizer lambda=0.2 (ablation) | adversarial contrastive learning (AERM) · CIFAR-10, FGSM eps=0.02 | adversarial accuracy: 45.93 -> 55.64 | Improved |
| add Frobenius-norm regularizer lambda=0.2 (ablation) | adversarial contrastive learning (AERM) · CIFAR-10, FGSM eps=0.02 | clean accuracy: 54.28 -> 66.68 | Improved |
| add Frobenius-norm regularizer lambda=0.002 (ablation) | adversarial contrastive learning (AERM) · CIFAR-10, PGD eps=0.01 | clean accuracy: 75.73 -> 74.8 | Hurt |
| add Frobenius-norm regularizer lambda=0.002 to 0.05 (ablation) | adversarial contrastive learning (AERM) · CIFAR-10, FGSM eps=0.02 | adversarial accuracy: 45.93 -> 45.94 | No clear effect |
| increase block size from 1 to 10 (ablation) | adversarial contrastive learning · CIFAR-10, PGD and FGSM | adversarial accuracy | Improved |
Closest papers by meaning
- On the Generalization Analysis of Adversarial Learning (ICML 2022 · similarity 0.90)
- Adversarial Rademacher Complexity of Deep Neural Networks (JMLR 2026 · similarity 0.89)
- Formulating Robustness Against Unforeseen Attacks (NeurIPS 2022 · similarity 0.89)
- Adversarial Training With Anti-Adversaries (TPAMI 2024 · similarity 0.89)
- Stability and Generalization of Adversarial Training for Shallow Neural Networks with Smooth Activation (NeurIPS 2024 · similarity 0.87)
- Generalization Analysis for Deep Contrastive Representation Learning (AAAI 2025 · 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.
