Extracting Alternative Solutions from Benders Decomposition

Abstract

We show how to extract alternative solutions for optimization problems solved by Benders Decomposition. In practice, alternative solutions provide useful insights for complex applications; some solvers do support generation of alternative solutions but none appear to support such generation when using Benders Decomposition. We propose a new post-processing method that extracts multiple optimal and near-optimal solutions using the cut-pool generated during Benders Decomposition. Further, we provide a geometric framework for understanding how the adaptive approximation in Benders Decomposition relates to alternative solutions. We demonstrate this technique on stochastic programming and interdiction modeling, and we highlight use cases that require the ability to enumerate all optimal solutions.