The question of the initial diversity in artificial evolutionary systems was addressed using gene expression programming. Due to the varied set of genetic operators and the high efficiency of the algorithm, it was possible to compare dissimilarly performing systems such as systems evolving under mutation alone and systems undergoing only recombination. As most existing artificial evolutionary systems rely either on mutation or recombination, this analysis can help understand the different evolutionary strategies followed by each system.
The results obtained in this work show that, on the one hand, systems using the high-performing mutation operator are not only more efficient but also capable of adaptation under extreme evolutionary bottlenecks. In fact, these systems show no correlation between the size of the founder population and success rate. Consequently, in the course of a run, this kind of system is never caught in evolutionary cul-de-sacs and, therefore, evolves without end.
On the other hand, systems relying on recombination alone not only perform poorly but also are unable to adapt and evolve when populations pass through a really tight bottleneck. Consequently, these systems not only are useless whenever only one viable individual is available to start an evolutionary epoch but also frequently become irrevocably stuck at evolutionary cul-de-sacs the system itself creates. Because of this, it is mandatory that these systems guarantee a high level of genetic diversity in initial populations for one thing, and for another, the population sizes in these systems must be huge in order to prevent evolutionary cul-de-sacs from happening. Obviously, these systems are highly expensive and impractical.