Abstract:

The common explanation of the manner in which genetic algorithms (GAs) process individuals in a population of contending solutions relies on the ``building block hypothesis.'' This suggests that successively better solutions are generated by combining useful parts of extant solutions. An alternative explanation is presented which focuses on the collective phenomena taking place in populations that undergo recombination. The new explanation is derived from investigations in evolution strategies (ESs). The principles studied are general, and hold for all evolutionary algorithms (EAs), including genetic algorithms (GAs). Further, they appear to be somewhat analogous to some theories and observations on the benefits of sex in biota.

 

Back to Publications
back to my Homepage