Absolute correlation between predictions on two groups of examples.

Inherits From: MinDiffLoss

name Name used for logging or tracking. Defaults to 'absolute_correlation_loss'.
enable_summary_histogram Optional bool indicating if tf.summary.histogram should be included within the loss. Defaults to True.

Absolute correlation measures how correlated predictions are with membership (regardless of direction). The metric guarantees that the result is 0 if and only if the two distributions it is comparing are indistinguishable.

The sensitive_group_labels input is used to determine whether each example is part of the sensitive group. This currently only supports hard membership of 0.0 or 1.0.

For more details, see the paper.