domain-anomaly-detection
Skill8 starsChanged 4 months agoMITView on GitHub
Use for anomaly, outlier, and novelty detection across data types: industrial or visual defect detection, time-series or sensor anomalies and predictive maintenance, tabular fraud or intrusion detection, and out-of-distribution detection. Usually one-class or unsupervised. Picks the method by data type, sets the threshold honestly, and uses imbalance-aware metrics. Triggers on 'anomaly detection', 'outlier', 'defect detection', 'novelty', 'fraud', 'intrusion', 'out of distribution', 'predictive maintenance', 'one-class'.
In this repository22 other files it gives its agents.
- accuracy-improvement-loopSkill
- data-rigor-and-leakageSkill
- dataset-profilingSkill
- deployment-explainabilitySkill
- domain-3dSkill
- domain-audio-speechSkill
- domain-computer-visionSkill
- domain-generativeSkill
- domain-graphSkill
- domain-medical-imagingSkill
- domain-multimodalSkill
- domain-nlp-llmSkill
- domain-recommenderSkill
- domain-reinforcement-learningSkill
- domain-tabularSkill
- domain-time-seriesSkill
- domain-videoSkill
- literature-reviewSkill
- ml-research-methodologySkill
- notebook-deliverySkill
- rigorous-evaluationSkill
- training-optimizationSkill
The file
Files
Anomaly Detection - Method Selection
Anomalies are rare and often unlabeled, so most setups are one-class: train on normal data, then flag deviations. Route by data type.
Decision table
| Data type | Recommended (recent) | Notes |
| Images or industrial defects | PatchCore (2022) or EfficientAD (WACV 2024) | few or no anomalies at train time. MVTec-style. Metric is image and pixel AUROC, and AUPRO. |
| Time-series or sensors | reconstruction or forecasting residual with a threshold; Anomaly Transformer as a deep option | read the metric caveat below. |
| Tabular (fraud, intrusion) | Isolation Forest and LOF baselines; deep methods (deep SVDD, autoencoder) when data is large | metric is AUROC and PR-AUC. |
| Graph or network | one-class or reconstruction-based GNN | fraud rings, intrusion. |
Cross-cutting practice
- Set the decision threshold on a validation set of known-normal data (plus a few anomalies if available), not on test.
- Imbalance is extreme, so use PR-AUC or AUROC, not accuracy.
- IMPORTANT time-series caveat: the common "point-adjustment" evaluation massively inflates F1 and is misleading. Report PR-AUC, range or affiliation-based metrics, or VUS, and state clearly whether point-adjustment was used.
- Explainability: reconstruction-error maps (images and time-series), SHAP or feature attribution (tabular), and which timestamp or region triggered the alert.
- Improve results: use
accuracy-improvement-loop; evaluate withrigorous-evaluation.
Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
No reviews yet
People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.
- It worked
- Partly
- It didn't
Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.
Your trail
Files you open will collect here, so you can find them again.
Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.
Now do this for your own.
Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.