Research map / Label-efficient and robust learning / OOD and anomaly detection
ADGym: Design Choices for Deep Anomaly Detection
Minqi Jiang, Chaochuan Hou, Ao Zheng, Songqiao Han, Hailiang Huang, Qingsong Wen, Xiyang Hu, Yue Zhao
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 pipelines with a meta-predictor instead of using the strongest fixed SOTA method | ADGym · Real-world datasets, n_a=5 | AUCROC: 0.784 -> 0.821 | Improved |
| Select pipelines with a meta-predictor instead of using the strongest fixed SOTA method | ADGym · Real-world datasets, n_a=10 | AUCROC: 0.816 -> 0.849 | Improved |
| Select pipelines with a meta-predictor instead of using the strongest fixed SOTA method | ADGym · Real-world datasets, n_a=20 | AUCROC: 0.843 -> 0.869 | Improved |
| Use supervised validation selection rather than random pipeline selection | ADGym · Real-world datasets, n_a=5 | AUCROC: 0.735 -> 0.613 | Hurt |
| Ensemble predicted pipelines rather than using a single DL meta-predictor (ablation) | ADGym DL meta-predictor · Real-world datasets, n_a=5 | AUCROC: 0.800 -> 0.813 | Improved |
| Ensemble predicted pipelines rather than using a single DL meta-predictor (ablation) | ADGym DL meta-predictor · Real-world datasets, n_a=10 | AUCROC: 0.836 -> 0.843 | Improved |
| Ensemble predicted pipelines rather than using a single DL meta-predictor (ablation) | ADGym DL meta-predictor · Real-world datasets, n_a=20 | AUCROC: 0.859 -> 0.865 | Improved |
| Expand the candidate design space (ablation) | ADGym meta-predictors · Real-world datasets, n_a=5 | AUCROC: 0.813 -> 0.829 | Improved |
| Expand the candidate design space (ablation) | ADGym meta-predictors · Real-world datasets, n_a=10 | AUCROC: 0.843 -> 0.853 | Improved |
| Expand the candidate design space (ablation) | ADGym meta-predictors · Real-world datasets, n_a=20 | AUCROC: 0.865 -> 0.876 | Improved |
| Refine the design space by removing choices below the training-set median (ablation) | ADGym meta-predictors · Real-world datasets | AUCROC | No clear effect |
| Use more complicated predictor losses instead of MSE (ablation) | ADGym meta-predictor · Real-world datasets | AUCROC | No clear effect |
Closest papers by meaning
- ADPretrain: Advancing Industrial Anomaly Detection via Anomaly Representation Pretraining (NeurIPS 2025 · similarity 0.86)
- We Need to Rethink Benchmarking in Anomaly Detection (NeurIPS 2026 · similarity 0.85)
- SEAD: Unsupervised Ensemble of Streaming Anomaly Detectors (ICML 2025 · similarity 0.85)
- AD-FM: Multimodal LLMs for Anomaly Detection via Multi-Stage Reasoning and Fine-Grained Reward Optimization (AAAI 2026 · similarity 0.85)
- ResAD: A Simple Framework for Class Generalizable Anomaly Detection (NeurIPS 2024 · similarity 0.85)
- Position: Quo Vadis, Unsupervised Time Series Anomaly Detection? (ICML 2024 · 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.
