Technical Note—Alternative Model Formulations in Nonlinear Programming—Some Disastrous Results
Abstract
This paper describes some experiments with two alternative formulations of a nonlinear programming model. One apparently suitable formulation proves to be numerically disastrous for several algorithms. The paper discusses the reasons for this behavior and argues that solution systems need to be strengthened so that they can cope with alternative model formulations. As argued in the paper, it would be desirable to improve the numerical algorithms themselves, or to develop computerized modeling systems that can reformulate models into the form most suitable for a particular existing algorithm.

