renate.shift.ks_detector module

class renate.shift.ks_detector.KolmogorovSmirnovCovariateShiftDetector(feature_extractor=None, batch_size=32, num_preprocessing_workers=0, device='cpu')

Bases: ShiftDetectorWithFeatureExtractor

A Kolmogorov-Smirnov (KS) test on each feature.

A KS test is a univariate two-sample test, which we perform separately for each feature. To aggregate these tests without running into multiple-testing problems, we use a Bonferroni correction, as proposed in

Rabanser et al. Failing Loudly: An Empirical Study of Methods for Detecting Dataset Shift. NeurIPS 2019.