Computational strategies for improved MINLP algorithms

Abstract: In order to improve the efficiency for solving MINLP problems, we present in this paper three computational strategies. These include multiple-generation cuts, hybrid methods and partial surrogate cuts for the Outer Approximation and Generalized Benders Decomposition. The properties and convergence of the strategies are analyzed. Five new MINLP algorithms are described based on the proposed strategies, and their implementation is discussed. Results of numerical experiments are reported for benchmark MINLP problems to demonstrate the efficiency of the proposed methods.