Applicability of Nonlinear Programming in Decision Support Systems: Realization of User-Friendly Software

摘要

To be able to use nonlinear programming algorithms for non-institutional decision support, it is necessary to put them into user-friendly software that handles the formulation and optimization of models. The formulation and optimization of mathematical models are sometimes separated. The interest here will be mainly in the optimization part. To be able to consider the 'optimization of a model' as a black box, the optimization part should provide options to let the choice of the appropriate algorithm be made automatically. For further support of the 'model-builder', diagnostic facilities should suggest weaknesses in the problem formulation. The first chapter of the paper gives an introduction to problems encountered with the application of nonlinear models. The remaining chapters describe how, in a partitioned algorithm, information gathered during the execution of the problems can be used to adapt parameter values for the algorithm, or even to decide to use a different algorithm starting from this point. Such an approach was followed in the development of a new system for nonlinear optimization. The algorithms in the system are based on algorithms for quadratic programming that were developed by Beale and Benveniste. (Copyright (c) 1987 by Faculty of Mathematics and Informatics, Delft, The Netherlands.)