The inverse Gomory corner relaxation (GCR) provides a tighter bound on the inverse integer programming (IP) optimal value than the inverse LP relaxation under mild conditions, and can be formulated as the inverse of a shortest path problem, leading to a more computationally tractable approach.
This paper proposes a novel inverse optimization method to learn the implicit convex constraints of an optimization problem from a set of expert-accepted and rejected solutions, aiming to improve the efficiency and accuracy of future decision-making processes.