Perturbed variations of penalty function methods. Example: Projective SUMT (Q2639773)

This article is motivated by the Projective SUMT (Sequential Unconstrained Minimization Technique) proposed originally by G. P. McCormick. The Projective SUMT is a penalty function technique for solving a certain convex programming problem, based on solving the differential equation associated with a barrier function minimizing trajectory. The author shows that the Projective SUMT is a simple variation of the classical SUMT and that all results (convexity and duality results, convergence, satisfaction of the Karush-Kuhn-Tucker conditions, etc.) that have been developed for the classical SUMT are applicable to the Projective SUMT with simple modifications. The author defines new classes of penalty functions by introducing simple linear perturbations of classical penalty functions or, equivalently, perturbations of the given convex programming problem. Further, the author indicates several classes of augmented penalty functions that he feels merit further study.