Examples

1. 1.

   For example, one might apply post-processing to a binary classification model by setting a classification threshold such that equality of opportunity is maintained for some attribute by checking that the true positive rate is the same for all values of that attribute.

Real-world uses

Created for this library

1. 1.

   An ML platform team applies post-processing to model outputs to apply business rules like minimum confidence thresholds.

2. 2.

   A search team applies post-processing such as diversity filtering after model scoring so a single page is not dominated by similar items.

3. 3.

   A risk team applies post-processing to model outputs to enforce policy caps before sending decisions to downstream systems.