A Hybrid Reinforcement and Self-Supervised Learning Aided Benders Decomposition Algorithm
For optimization practitioners, this work offers a faster method for solving mixed integer nonlinear programs, though it is evaluated on a single case study and may be incremental.
The paper proposes a hybrid reinforcement and self-supervised learning framework to accelerate generalized Benders decomposition, achieving a 57.5% reduction in solution time while consistently recovering optimal solutions on a mixed integer nonlinear programming case study.
We propose a hybrid reinforcement and self-supervised learning framework for accelerating generalized Benders decomposition (GBD). In this framework, a graph based reinforcement learning agent operates on a bipartite representation of the master problem and, together with a verification mechanism, determines the integer variable assignments that solve the master problem. These assignments are then used as inputs to a KKT informed neural network, trained via self supervision to predict primal dual solutions that approximately satisfy the Karush Kuhn Tucker conditions of the subproblem. The predicted solutions are used to construct Benders cuts directly. The framework is evaluated on a mixed integer nonlinear programming case study, where it achieves a 57.5% reduction in solution time relative to classical GBD while consistently recovering optimal solutions across all test instances.