Practical use of generalized simulated annealing optimization on microcomputers

Abstract

The practical use of Generalized Simulated Annealing (GSA) optimization is explored to determine additional benefits beyond the global minima location capability. In particular a means of using the strengths of both GSA and Damped Least Squares (DLS) optimization techniques is examined in order to increase the amount of usable information and to enhance usefulness in a microcomputer environment. 1. BACKGROUND The optimization ofthe performanceofan opticalsystem is most often performed bydamped leastsquares (DLS) techniques. This methodand its variants has proven to beeffective for the reduction ofsystem merit functions overa broad range ofproblem types with quite rapid convergence rates. The effectiveness of DLS optimization will often depend on the quality of the starting point the construction of the merit function and the proximity of the starting point to local minima ofthe merit function. The characteristic of the DLS technique to compute an optimum solution to minimize the merit function based on the instantaneous first and in some cases second partial derivatives gives rise to the technique''s strongest and weakest points. The strongest point is its rapid convergence while the weakest point is the inability to escape local minima in order to locate a superior local minima or the global minima. The strength of rapid convergencewas important when the majority oflens design programs where on a mainframe computers were processing time is costly. The inability to escape local minima weakness was most often overcome