Research map / Adversarial robustness and security / Certified and empirical robustness
Adversarially Robust One-class Novelty Detection
Shao-Yuan Lo, Poojan Oza, Vishal M. Patel
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 Vector-PCA (principal latent vector replacement) (ablation) | PGD-AT autoencoder · MNIST, PGD attack | mAUROC: 0.357 -> 0.566 | Improved |
| add Vector-PCA (principal latent vector replacement) (ablation) | PGD-AT autoencoder · CIFAR-10, PGD attack | mAUROC: 0.145 -> 0.215 | Improved |
| add feature denoising block after Vector-PCA (ablation) | Vector-PCA · MNIST, PGD attack | mAUROC: 0.566 -> 0.582 | Improved |
| add feature denoising block after Vector-PCA (ablation) | Vector-PCA · CIFAR-10, PGD attack | mAUROC: 0.215 -> 0.215 | No clear effect |
| add Spatial-PCA instead of feature denoising (ablation) | Vector-PCA+FD · MNIST, PGD attack | mAUROC: 0.582 -> 0.706 | Improved |
| add Spatial-PCA instead of feature denoising (ablation) | Vector-PCA+FD · F-MNIST, PGD attack | mAUROC: 0.505 -> 0.613 | Improved |
| increase k_V from 1 to 64 (ablation) | PrincipaLS · MNIST, PGD attack | mAUROC: 0.706 -> 0.557 | Hurt |
| increase k_V from 1 to 4 (ablation) | PrincipaLS · MNIST, clean | mAUROC: 0.973 -> 0.975 | Mixed |
| use k_S=1 instead of 8 (ablation) | PrincipaLS · MNIST, PGD attack | mAUROC: 0.706 -> 0.549 | Hurt |
| attach PrincipaLS to ARAE | ARAE · MNIST, PGD attack | mAUROC: 0.133 -> 0.695 | Improved |
| attach PrincipaLS to Puzzle-AE | Puzzle-AE · MNIST, PGD attack | mAUROC: 0.295 -> 0.599 | Improved |
| add PrincipaLS to naturally trained AE | AE · MNIST, clean | mAUROC: 0.964 -> 0.973 | Improved |
Closest papers by meaning
- OCGAN: One-Class Novelty Detection Using GANs With Constrained Latent Representations (CVPR 2019 · similarity 0.92)
- Adversarially Learned One-Class Classifier for Novelty Detection (CVPR 2018 · similarity 0.91)
- Generative Probabilistic Novelty Detection with Adversarial Autoencoders (NeurIPS 2018 · similarity 0.89)
- Multiple Class Novelty Detection Under Data Distribution Shift (ECCV 2020 · similarity 0.89)
- Old Is Gold: Redefining the Adversarially Learned One-Class Classifier Training Paradigm (CVPR 2020 · similarity 0.88)
- Universal Novelty Detection Through Adaptive Contrastive Learning (CVPR 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.
