Skip to content

DataEval documentation

dataeval.shift

dataeval.shift¶

Detect changes in data between different datasets.

Submodules¶

`update_strategies`	Update strategies inform how the drift detector classes update the reference data.

Classes¶

`ChunkResult`	Result for a single chunk in chunked drift detection.
`DriftDomainClassifier`	Multivariate Domain Classifier based drift detector.
`DriftDomainClassifierStats`	Statistics from Multivariate Domain Classifier drift detection.
`DriftKNeighbors`	K-Nearest Neighbors based drift detector.
`DriftKNeighborsStats`	Statistics from K-Nearest Neighbors drift detection.
`DriftMMD`	Drift detector using Maximum Mean Discrepancy (MMD) Drift Detection with permutation test.
`DriftMMDStats`	Statistics from MMD permutation test.
`DriftReconstruction`	Reconstruction-based drift detector using autoencoders.
`DriftReconstructionStats`	Statistics from reconstruction-based drift detection.
`DriftUnivariate`	Drift detector using univariate statistical tests.
`DriftUnivariateStats`	Per-feature statistics from univariate drift detection.
`OODDomainClassifier`	Domain Classifier based Out-of-Distribution detector.
`OODKNeighbors`	K-Nearest Neighbors Out-of-Distribution detector.
`OODReconstruction`	Autoencoder (AE) or Variational Autoencoder (VAE) based out-of-distribution detector.

Output Classes¶

`DriftOutput`	Output class for drift detector classes.
`OODOutput`	Output class for predictions from out-of-distribution detectors.
`OODScoreOutput`	Output class for instance and feature scores from out-of-distribution detectors.