RiskBoundsandCalibrationforaSmart Predict-then-OptimizeMethod
–Neural Information Processing Systems
Moreover, since the SPO loss is not continuous nor convex in general [Elmachtoub and Grigas, 2021], which makesthe training ofaprediction model computationally intractable, Elmachtoub and Grigas [2021] introduced a novel convex surrogate loss, referred to as the SPO+ loss.
Neural Information Processing Systems
Feb-10-2026, 21:33:22 GMT