dataeval.metrics.bias#

Bias metrics check for skewed or imbalanced datasets and incomplete feature representation which may impact model performance.

Classes#

`BalanceOutput`	Output class for `balance()` bias metric
`CoverageOutput`	Output class for `coverage()` bias metric
`DiversityOutput`	Output class for `diversity()` bias metric
`ParityOutput`	Output class for `parity()` and `label_parity()` bias metrics

`balance`(metadata[, num_neighbors])	Mutual information (MI) between factors (class label, metadata, label/image properties)
`coverage`(embeddings[, radius_type, k, percent])	Class for evaluating coverage and identifying images/samples that are in undercovered regions.
`diversity`(metadata[, method])	Compute diversity and classwise diversity for discrete/categorical variables and,
`label_parity`(expected_labels, observed_labels[, ...])	Calculate the chi-square statistic to assess the parity between expected and
`parity`(metadata)	Calculate chi-square statistics to assess the linear relationship between multiple factors