Each component algorithm retains and uses its recommended set
of parameters, which is the standard practice in the evolutionary
community. No parameter tuning and control is required. Instead,
our approach expects individual algorithms to play
complementary roles. Different algorithms which excel in
different problems will stand out when required. Moreover, as
different algorithms may be the best for different computationalbudget in the sense of absolute maximum number of fitness
evaluations, the approach fully allows the selection of the best
algorithm given a fixed computational budget, as well as
automatic algorithm switching as the budget varies.