Necessary and sufficient conditions of unacceptable solutions in NLP discriminant analysis

Over the last ten years, there has been a growing interest in the application of mathematical programming techniques to discriminant analysis. Most research so far has focused on linear programming models. While many significant results in LP discriminant analysis have been obtained, little has been done on nonlinear programming models due to their complexity. This paper establishes necessary and sufficient conditions of unacceptable solutions for a number of nonlinear programming discriminant models.