 Loss function

A loss function assigns a numerical penalty to a prediction and its observed outcome. Its expected value is the population risk optimized by a statistical learning procedure.