SUMT WITHOUT PARAMETERS.
Systems research memo.,
NORTHWESTERN UNIV EVANSTON ILL TECHNOLOGICAL INST
Pagination or Media Count:
The paper gives a variation of the Sequential unconstrained minimization technique SUMT for nonlinear programming. A new penalty function is introduced that can provide the required transformation for converting the original constrained problem into a sequency of unconstrained problems. The new feature is essentially that the objective function is treated as a moving constraint. Proof of convergence is given. The relationship between the new function and the one presently utilized in SUMT is specified in detail. The new function has a number of advantages such as the fact that it is free of dependence on parameters and is such as to force a decrease in the functional being minimized, from the given initial feasible point. Preliminary computational results, using an algorithm based on the function, indicated correct convergence. To date, however, the technique has not been tested adequately to draw any firm conclusions as regards its computational efficacy as compared to the present version of SUMT. Author