Use-case gallery

The use-case gallery is organized by problem theme. This is usually the best way to enter the documentation: choose the topic that looks like your application, then open a card with plots, captured output, and interpretation.

Start here

If you are new to the package, start with one of these examples:

Fraud and security

Suspicious tabular records, network traffic anomalies, and robust distance ranking.

Finance and risk

Heavy-tailed covariance, portfolio stress, and market-risk style examples.

Real ML datasets

Breast cancer, digits, and wine examples with sklearn baseline comparisons.

Scatter geometry

Compare covariance and scatter matrices with affine-invariant and log-Euclidean geometry.

Kernels
GP metrics

Similarity and kernels

Use robust scatter estimates as input metrics for kernels, GP workflows, and similarity methods.

Browse by topic

Finance
risk

Finance and risk

Portfolio covariance, stress monitoring, and heavy-tailed returns.

Fraud
security

Fraud, security, and networks

Fraud-like screening and network-flow anomaly examples.

Sensors
quality

Sensors and quality control

Sensor anomaly, predictive maintenance, and process monitoring.

Signals
embeddings

Biomedical, image, and embedding data

Feature-vector anomaly detection for signals, images, and embeddings.

Real ML
datasets

Real ML datasets

Reproducible built-in datasets with baseline metrics and plots.

ML
preprocess

Robust ML preprocessing

Use robust distances before downstream classification.

SPD
geometry

Scatter geometry

SPD matrix distances, geodesics, covariance drift, and robust shape comparison.

Kernels
similarity

Similarity, kernels, and GP metrics

Robust input metrics for kernel methods, Gaussian-process workflows, and embedding retrieval.

Which topic should I open?

Topic guide

Problem you have	Open this topic	Good first estimator
Fraud, suspicious transactions, unusual records	Fraud, security, and networks	FastMCD or AutoRobustAnomalyDetector
Portfolio, returns, covariance, stress periods	Finance and risk	RegularizedCauchy
Sensors, process drift, industrial faults	Sensors and quality control	FastMCD or RegularizedCauchy
Signal/image/embedding feature vectors	Biomedical, image, and embedding data	AutoRobustScatter or RegularizedCauchy
Reproducible ML benchmark examples	Real ML datasets	RobustOutlierDetector with baseline comparison
Clean training data before a classifier	Robust ML preprocessing	robust-distance filtering
Compare covariance/scatter matrices or monitor covariance drift	Scatter geometry	robustcov.geometry with fitted scatter estimates
Build robust similarity, kernel, or GP input metrics	Similarity, kernels, and GP metrics	robust input metric from RegularizedCauchy or FastMCD

Run the gallery

python examples/run_use_case_gallery.py

Run every gallery script:

python examples/run_use_case_gallery.py --all

Regenerate documentation assets

The gallery pages embed captured outputs and plots. Refresh them after changing examples:

python docs/generate_gallery_assets.py
sphinx-build -b html docs docs/_build/html

Topic pages

• Finance and risk
  • How to use this topic
  • Detailed pages
• Fraud, security, and network anomalies
  • How to use this topic
  • Detailed pages
• Sensors, industrial monitoring, and quality control
  • How to use this topic
  • Detailed pages
• Biomedical, image, and embedding data
  • How to use this topic
  • Detailed pages
• Real ML datasets
  • How to use this topic
  • Detailed pages
• Robust ML preprocessing
  • How to use this topic
  • Detailed pages

All detailed pages

• Finance-style heavy-tail covariance
• Portfolio covariance stress comparison
• Fraud-style tabular anomaly screening
• Network-traffic anomaly simulation
• Sensor anomaly detection
• Predictive-maintenance monitoring
• Quality-control monitoring
• Biomedical / signal-window anomaly detection
• Image-feature one-class anomaly detection
• Text / embedding outlier screening
• Breast-cancer screening as anomaly ranking
• Digits one-class anomaly detection
• Wine class screening
• Robust preprocessing before classification
• SPD geometry utilities
• Robust GP kernel / input metric
• Robust embedding retrieval / RAG filtering
• Multimodal anomaly detection