ABSTRACT
This paper explores acceleration of the harmonic mean algorithm for polynomial geometric programming using sensitivity analysis procedures. This acceleration is accomplished by using the sensitivity analysis methods, developed for ordinary geometric programs, to skip iterations of the harmonic algorithm. Several implementations of the acceleration procedure based on the size of the sensitivity analysis steps, frequency of resolving the harmonic program, and use of incremental procedure as a means of controlling step size are described and tested on a set of problems. The comparison of the implementations is based on the criteria of computational expense and accuracy of the solutions.
