next up previous

4.2.9 GA Algorithm Performance     continued...

Within the GA each control variable was represented by a 30 bit binary number (chromosome), or equivalently in decimal integers between and 0. So, when rescaled into real numbers with a range of 3, this representation gives quantization errors of just .

Figure 22 shows the progress in reducing the objective function for the same search. Both the fitness of the best individual within each population and the population average fitness are shown (note that the scales are different). These are the two standard measures of progress in a GA run. The difference between these two measures is indicative of the degree of convergence in the population.

Figure 22: Minimization of the Two-Dimensional Rosenbrock Function by a Genetic Algorithm --- Population Distribution of the First, Tenth,Twentieth, and Thirtieth Generations.