Evolvability in Developmental Systems

The developmental mapping from genotype to phenotype is responsible for much of the evolvability (adaptive varia-tion) exhibited in nature (Raff 1996; Kirschner & Gerhart 1998; Wagner & Altenberg 1996). Random mutations are transformed into structured phenotypic variation and the ef-fects of deleterious mutations are mitigated. This mech-anism is powerful precisely because search becomes con-strained, generating only highly adaptive phenotypes. In evolutionary computation, acquiring such constraints and bias is akin to learning the underlying structure of a par-ticular fitness function; such structure can be exploited to improve search efficiency. For example, when evolving a design for a coffee table, an evolvable encoding would dis-cover that table height and surface area correspond to fun-damental axes of variation, constraining search to solution