Prediction models for linear program right-hand sides are trained via decision error minimization and historical primal-dual solutions to ensure the true optimal solution remains feasible and optimal under the predicted constraints.
Two-stage predict+ optimize for milps with unknown parameters in constraints.Advances in Neural Information Processing Systems, 36:14247–14272, 2023
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
math.OC 1years
2026 1verdicts
UNVERDICTED 1representative citing papers
citing papers explorer
-
Decision-Aware Predictions for Right-Hand Side Parameters in Linear Programs
Prediction models for linear program right-hand sides are trained via decision error minimization and historical primal-dual solutions to ensure the true optimal solution remains feasible and optimal under the predicted constraints.