Ethics & Safety Intermediate
calibration layer
A post-prediction adjustment, typically to account for prediction bias.
Plain English Explanation
A post-prediction adjustment, typically to account for prediction bias. The adjusted predictions and probabilities should match the distribution of an observed set of labels.
How is it used?
Practitioners refer to calibration layer when building, training, or evaluating machine learning systems. It appears in research papers, product documentation, and technical discussions about AI capabilities and limitations.