Research map / Adversarial robustness and security / Certified and empirical robustness
API-Net: Robust Generative Classifier via a Single Discriminator
Xinshuai Dong, Hong Liu, Rongrong Ji, Liujuan Cao, Qixiang Ye, Jianzhuang Liu, Qi Tian
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 API-Net training schedule on top of Madry-initialized API-Net (ablation) | API-Net · CIFAR-10 | PGD-40 accuracy: 52.08 -> 63.13 | Improved |
| add API-Net training schedule on top of Madry-initialized API-Net (ablation) | API-Net · SVHN | PGD-40 accuracy: 58.58 -> 74.36 | Improved |
| apply API inference to off-the-shelf robust discriminators without training (ablation) | Madry/Trades models · CIFAR-10 | PGD-40 accuracy: approximate 5% | Improved |
| apply API inference to off-the-shelf robust discriminators without training (ablation) | Madry/Trades models · SVHN | PGD-40 accuracy: approximate 4% | Improved |
| initialize Trades from Madry model (lr 0.1) (ablation) | Trades · CIFAR-10 | PGD-40 accuracy: 52.08 -> 50.08 | Hurt |
| initialize ME-Net from Madry model (lr 0.1) (ablation) | ME-Net · CIFAR-10 | PGD-40 accuracy: 55.40 -> 52.92 | Hurt |
| lower learning rate to 0.01 when initializing Trades from Madry (ablation) | Trades · CIFAR-10 | PGD-40 accuracy: 50.08 -> 47.24 | Hurt |
| raise eps_ap from 6/255 to 14/255 (ablation) | API-Net · CIFAR-10 | PGD-40 accuracy: 55.09 -> 63.13 | Improved |
| raise eps_ap from 6/255 to 12/255 (ablation) | API-Net · SVHN | PGD-40 accuracy: 61.14 -> 74.40 | Improved |
| raise eps_ap from 12/255 to 14/255 (ablation) | API-Net · MNIST | PGD-40 accuracy: 97.12 -> 96.32 | Hurt |
| raise eps_ap from 12/255 to 16/255 (ablation) | API-Net · SVHN | PGD-40 accuracy: 74.40 -> 73.39 | Hurt |
| replace Trades with API-Net on same ResNet18 | API-Net vs Trades · CIFAR-10 | PGD-40 accuracy: 52.08 -> 63.13 | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- ME-Net — ME-Net: Towards Effective Adversarial Robustness with Matrix Estimation (ICML 2019) · better on 3 of 3 shared results
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- Robust Inference via Generative Classifiers for Handling Noisy Labels (ICML 2019 · similarity 0.88)
- Understanding the Limitations of Conditional Generative Models (ICLR 2020 · similarity 0.88)
- Adversarial vulnerability for any classifier (NeurIPS 2018 · similarity 0.87)
- Wide Two-Layer Networks can Learn from Adversarial Perturbations (NeurIPS 2024 · similarity 0.86)
- Generative Classifiers Avoid Shortcut Solutions (ICLR 2025 · similarity 0.86)
- Fusing Conditional Submodular GAN and Programmatic Weak Supervision (AAAI 2024 · 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.
