Performance
The sixteen optimization problems presented were solved with GTA using
500 cyclists and considering 1000 as the maximum number of iterations
allowed. Figure 2 shows the boxplot analysis for the 100 optimizations
performed. In most of the cases the deviation is so low that it is not
possible to observe the height of the boxplot in the figure. For the
Rosenbrock and Dixon Price functions a local optimal solution was
invariably found, but the minimum value was never reached. The number of
evaluations of the objective function for all the cases was, in average,
high, close to the maximum of 500,000. However, since no improvement
tolerance was used to stop the optimization, after just a few iterations
the gains soon become negligible, as can be seen in Figure 3. Thus, the
three fundamental features of an optimization procedure, namely, few
parameters to adjust, fast convergence, and reliability of the results,
are satisfied for the studied cases.