Figure 3

Comparison of the standard optimization algorithms applied to each deterministic model. Five algorithms show outstanding results on the reversible models. These are binGA, cmaESplus, DE, PSO, and Tribes. On the irreversible models, binGA does not yield such good average results, but is still capable of finding good local optima. For these models, cmaESplus and Tribes also find pre-eminent results, whereas PSO and DE are among the best algorithms, except for the GMM models. The real-valued GA performs worst in all cases. The hill climbers show differences in their effectiveness when invoked with various numbers of multi-starts but cannot compete with the other algorithms. The term "fitness" is used to define a quality measure of possible solutions, which is minimized.