Article ID: | iaor1993691 |
Country: | Netherlands |
Volume: | 12 |
Issue: | 1 |
Start Page Number: | 53 |
End Page Number: | 58 |
Publication Date: | Jul 1992 |
Journal: | Operations Research Letters |
Authors: | Gulledge Thomas R., Ringuest Jeffrey L. |
Keywords: | programming: multiple criteria |
In recent years there has been much interest in developing interactive algorithms for solving multiple objective optimization problems. In this paper the authors present an approach for estimating the gradient within a line-search algorithm. The approach uses a numerical approximation to the gradient obtained by fitting a first-order experimental desgin to the decision maker’s value function. Test results indicate that the algorithm provides accurate utility predictions for numerically assessed gradient vectors, relative to analytical gradient vectors.