A New Class of Algorithms for Multipoint Network Optimization

Abstract

A class of second-order greedy algorithms (SOGA) for the optimization of multipoint networks is defined that is capable of generating improved solutions to problems by an average of more than two percent, relative to the solutions generated by the Esau-Williams algorithm. These SOGA utilize repeated calls of a modified Esau-Williams procedure. They consequently take much longer to execute than does the Esau-Williams algorithm. Nevertheless, problems involving up to 400 points can be solved on a large computer, and 200-point problems can be solved on machines of intermediate size.