On Optimization Model Formulation Using Neural Networks
Paweł Gąsiorowski · 2002
Mathematical programming is a widely used tool in Operational Research. The first — and usually the most difficult — step of problem solving requires formulation of the model: finding the mathematical description of the relationships between decisions and their outcomes. The researcher is faced with the problem of finding the proper functions and their parameters. This kind of problems is dealt with using approximation theory. The paper presents theorems and assumptions for using neural networks as approximators of functions in formulation of the optimization models Thanks to neural networks’ approximation capabilities the difficulties connected with finding the functional forms of relationships and their parameters can be avoided. The method of identification of parameters is illustrated with an exemplary application in the area of bank credit portfolio optimization. These keywords were added by machine and not by the authors. This process is experimental and the keywords may be updated as the learning algorithm improves.