Research map / Probabilistic and scientific modeling / Variational and generative sampling
Noise-Robust Density Estimation for Tabular Data Anomaly Detection
Dazhi Fu, Zhao Zhang, Jicong Fan
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 |
|---|---|---|---|
| Select smallest-Jacobian-norm sources as data sources instead of largest (variance assumption) (ablation) | NRDE · WPBC | AUROC (%): 59.5 -> 64.4 | Improved |
| Select smallest-Jacobian-norm sources as data sources instead of largest (variance assumption) (ablation) | NRDE · Cardiotocography | AUROC (%): 76.5 -> 87.1 | Improved |
| Select smallest-Jacobian-norm sources as data sources instead of largest (variance assumption) (ablation) | NRDE · vertebral | AUROC (%): 70.1 -> 82.7 | Improved |
| Select smallest-Jacobian-norm sources as data sources instead of largest (variance assumption) (ablation) | NRDE · Waveform | AUROC (%): 72.1 -> 86.7 | Improved |
| Select smallest-Jacobian-norm sources as data sources instead of largest (variance assumption) (ablation) | NRDE · Speech | AUROC (%): 51.7 -> 58.6 | Improved |
| Increase data-to-noise variance ratio from 0.5 to 9 (ablation) | NRDE · Synthetic data | AUROC (%): 65.8 -> 87.3 | Improved |
| Use true number of data sources (oracle) instead of estimating | NRDE · Synthetic data | AUROC (%): 80.9 -> 86.0 | Improved |
| Use NRDE-generated pure data for KNN training and testing | kNN · vertebral | AUROC (%): +22.7 | Improved |
| Use NRDE-generated pure data for ECOD training and testing | ECOD · Wilt | AUROC (%): +30.5 | Improved |
| Use NRDE-generated pure data for KDE training and testing | KDE · annthyroid | AUROC (%): +3.6 | Improved |
| Use NRDE-generated pure data for KNN | kNN · landsat | AUROC (%): +0.4 | Improved |
| Replace RealNVP with Jacobian-regularized flow (NRDE) | RealNVP · ADBench 47 datasets | AUROC (%): 82.3 -> 88.5 | Improved |
Closest papers by meaning
- Unsupervised Anomaly Detection for Tabular Data Using Deep Noise Evaluation (AAAI 2025 · similarity 0.90)
- Beyond Individual Input for Deep Anomaly Detection on Tabular Data (ICML 2024 · similarity 0.88)
- Tuning-Free One-Class Discriminant Learning for Tabular Anomaly Detection (ICML 2026 · similarity 0.88)
- Tab-Shapley: Identifying Top-k Tabular Data Quality Insights (AAAI 2025 · similarity 0.86)
- Disentangling Tabular Data Towards Better One-Class Anomaly Detection (AAAI 2025 · similarity 0.85)
- SEAD: Unsupervised Ensemble of Streaming Anomaly Detectors (ICML 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.
