Abstract

The use of Generalized Simulated Annealing algorithms has been studied by several authors for its potential value to overcome the traditional shortcomings of damped least squares optimization. Damped least squares is likely to converge to local minima and once found the method cannot escape this condition by itself. As a result the effectiveness of this type of optimization is strongly linked to the qualities of the starting point, the merit function definition, and the selection of variables.

© 1990 Optical Society of America