Research map / Label-efficient and robust learning / OOD and anomaly detection
Supervised Anomaly Detection via Conditional Generative Adversarial Network and Ensemble Active Learning
Zhi Chen, Jiang Duan, Kang Li, Guoping Qiu
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 multiple discriminators (ensemble of up to 21) instead of one (ablation) | EAL-GAN · 20 real-world datasets | AUC and Gmean | Improved |
| add ensemble learning loss with modulating factors instead of standard adversarial loss (ablation) | EAL-GAN · synthetic datasets | AUC and Gmean | Improved |
| add embedding layer (projection discriminator) to discriminator (ablation) | EAL-GAN · synthetic datasets | AUC and Gmean | Improved |
| add ensemble active sampling instead of random sampling (ablation) | EAL-GAN · synthetic datasets | AUC and Gmean | Improved |
| use one ensemble of discriminators instead of a single discriminator (ablation) | EAL-GAN · synthetic single-cluster data | AUC and Gmean | Improved |
| train detector on real plus generated data instead of generated data only (ablation) | EAL-GAN · 20 real-world datasets | AUC: 0.906 -> 0.938 | Improved |
| train detector on generated data only versus real data only (ablation) | EAL-GAN · 20 real-world datasets | AUC: 0.911 vs 0.906 | No clear effect |
| increase Fake:Real ratio beyond 100:1 (ablation) | EAL-GAN · 20 real-world datasets | AUC: 0.958 -> 0.955 | Hurt |
| increase labeled sampling ratio from 5% to 95% (ablation) | EAL-GAN · 20 real-world datasets | AUC: 0.951 -> 0.977 | Improved |
| use 4-layer generator instead of 3-layer or 5-layer (ablation) | EAL-GAN · 20 real-world datasets | AUC: 0.9% over 3-layer, 0.5% over 5-layer | Improved |
| use EAL-GAN instead of unsupervised baselines on large datasets | EAL-GAN vs CBLOF · large datasets (12) | AUC gain (%): +10.24 -> +20.66 | Improved |
| compare EAL-GAN to best deep baseline DevNet | EAL-GAN · 20 real-world datasets | average AUC: 0.900 -> 0.951 | Improved |
Closest papers by meaning
- GAN Ensemble for Anomaly Detection (AAAI 2021 · similarity 0.93)
- Old Is Gold: Redefining the Adversarially Learned One-Class Classifier Training Paradigm (CVPR 2020 · similarity 0.90)
- Contamination-Resilient Anomaly Detection via Adversarial Learning on Partially-Observed Normal and Anomalous Data (ICML 2024 · similarity 0.86)
- Scalable, Explainable and Provably Robust Anomaly Detection with One-Step Flow Matching (NeurIPS 2025 · similarity 0.86)
- ADPretrain: Advancing Industrial Anomaly Detection via Anomaly Representation Pretraining (NeurIPS 2025 · similarity 0.84)
- PatchGuard: Adversarially Robust Anomaly Detection and Localization through Vision Transformers and Pseudo Anomalies (CVPR 2025 · similarity 0.84)
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.
