The development of methods to solve mixed-integer nonlinear programming (MINLP) problems has given rise to new solvers and improved the current ones. In this Article, the focus is set on the MINLP solvers in the General Algebraic Modeling System (GAMS) and especially GAMS/AlphaECP. In this Article, a comprehensive comparison of the MINLP solvers is made. In June 2007, GAMS introduced new MINLP solvers: CoinBonmin, AlphaECP, and LINDOGlobal. A description of the improvements made in the second half of 2007 to AlphaECP is given, and the performance of the MINLP solvers in GAMS is examined. Furthermore, the performance of AlphaECP for both MINLP and nonlinear programming (NLP) problems is studied, in conjunction with different subsolvers. Two large collections of problems, MINLPLib and GLOBALLib, are used for all solver performance comparisons.

No abstract

No abstract

No abstract