Time-Quality Framework for a Theory-Based Analysis and Design of Representations
Franz Rothlauf · Studies in fuzziness and soft computing · 2002
Over the last decades, researchers gained more and more knowledge about the principles of genetic and evolutionary algorithms (GEAs) and were able to formulate a theory describing the behavior of GEAs more precisely (Goldberg, 1989c; Bäck, Fogel, & Michalewicz, 1997). The existing elements of GEA theory explain quite accurately the influence of many important GEA parameters, as well as selection, recombination, or mutation methods on the performance of GEAs. By using the existing GEA theory, straight forward design and the development of new, competent GAs (Goldberg, Deb, Kargupta, &Harik, 1993; Hank & Goldberg, 1996; Pelikan, Goldberg, & Cantú-Paz, 1999; Pelikan, Goldberg, & Lobo, 1999) became possible. However, concerning representations for GEAs, a framework which describes the influence of representations on the performance of GEAs is still missing, although it is well known that the used representation has a strong influence on GEA performance. Such a framework could help us to develop new representations in a more theory-guided manner and would be an important step towards a general theory of representations for GEAs.