Expand description
Distribution drift detection algorithms
Provides a DriftDetector trait and four concrete implementations:
| Detector | Method | Multivariate? |
|---|---|---|
KolmogorovSmirnovDetector | Two-sample KS test | No (1D) |
PopulationStabilityIndexDetector | PSI via binning | No (1D) |
WassersteinDetector | Earth-mover / Wasserstein-1 | No (1D) |
MaximumMeanDiscrepancyDetector | Kernel MMD² | Yes |
Each detector compares a reference window against a test window and
produces a DriftResult indicating whether drift was detected.
Structs§
- Drift
Result - Result of a drift detection test.
- Kolmogorov
Smirnov Detector - Two-sample Kolmogorov-Smirnov test for distribution shift.
- Maximum
Mean Discrepancy Detector - Maximum Mean Discrepancy (MMD) for multivariate drift detection.
- Population
Stability Index Detector - Population Stability Index (PSI) for measuring distribution shift.
- Wasserstein
Detector - Wasserstein-1 (earth mover’s) distance for 1-D distribution shift detection.
Traits§
- Drift
Detector - Trait for drift detectors that compare two 1-D sample arrays.