Sensitivity Analysis for Parametric Non-Linear Programming Using Penalty Methods.
GEORGE WASHINGTON UNIV WASHINGTON D C PROGRAM IN LOGISTICS
Pagination or Media Count:
Recently, it has been shown that a class of penalty function algorithms can readily be adapted to generate sensitivity analysis information for a large class of parametric nonlinear programming problems. In particular, estimates of the partial derivatives with respect to the problem parameters of the components of a solution vector and the optimal value function have been successfully calculated for a number of nontrivial examples. The approach has been implemented using the well known Sequential Unconstrained Minimization Technique SUMT computer program. This paper, a continuation and amplification of a recent paper by Armacost, gives a detailed summary of the significant underlying theoretical results, reviews recent additions to the computer program that include Lagrange multiplier sensitivity calculations, and elaborates on the kind of information that can be generated by further analyzing and interpreting results obtained in applying the techique to a well known inventory model. Author
- Theoretical Mathematics
- Computer Programming and Software