Accelerating Generalized Benders Decomposition for Wireless Resource Allocation

Generalized Benders decomposition (GBD) is a globally optimal algorithm for mixed integer nonlinear programming (MINLP) problems, which are NP-hard and can be widely found in the area of wireless resource allocation. The main idea GBD decomposing an MINLP problem into primal master problem, iteratively solved until their solutions converge. However, direct implementation time- memory-consuming. bottleneck high complexity increases over iterations. Therefore, we propose to leverage machine learning (ML) techniques accelerate aiming at decreasing problem. Specifically, utilize two different ML techniques, classification regression, deal with this acceleration task. In way, cut classifier regressor learned, respectively, distinguish between useful useless cuts. Only cuts added thus reduced. By using allocation device-to-device communication networks as example, validate that proposed method reduce computational without loss optimality has good generalization ability. applicable solving various problems since designs invariant problems.