GENETIC ALGORITHMS IN WAVELENGTH SELECTION - A COMPARATIVE-STUDY

This paper presents a comparative study involving a genetic algorithm, simulated annealing, and stepwise elimination, as methods for wavelength selection in multi-component analysis. The wavelength selection criteria used are the selectivity and accuracy after Lorber, and the minimal mean squared error after Sasaki. The genetic algorithm generally performed best. Stepwise elimination performed surprisingly good despite its local search heuristic. Simulated annealing performed worst, which is remarkable in view of the fact that this method is widely praised in the literature for properties similar to those of genetic algorithms, e.g., a probabilistic, non-local search heuristic.