Evolvability

Evolvability is defined as the capacity of a system for adaptive evolution. Evolvability is the ability of a population of organisms to not merely generate genetic diversity, but to generate adaptive genetic diversity, and thereby evolve through natural selection.

In order for a biological organism to evolve by natural selection, there must be a certain minimum probability that new, heritable variants are beneficial. Random mutations, unless they occur in DNA sequences with no function, are expected to be mostly detrimental. Beneficial mutations are always rare, but if they are too rare, then adaptation cannot occur. Early failed efforts to evolve computer programs by random mutation and selection showed that evolvability is not a given, but depends on the representation of the program. Analogously, the evolvability of organisms depends on their genotype-phenotype map. This means that biological genomes are structured in ways that make beneficial changes less unlikely than they would otherwise be. This has been taken as evidence that evolution has created not just fitter organisms, but populations of organisms that are better able to evolve.

Evolvability (computer science)

The term evolvability is used for a recent framework of computational learning introduced by Leslie Valiant in his paper of the same name and described below. The aim of this theory is to model biological evolution and categorize which types of mechanisms are evolvable. Evolution is an extension of PAC learning and learning from statistical queries.

General Framework

Let and be collections of functions on variables. Given an ideal function , the goal is to find by local search a representation that closely approximates . This closeness is measured by the performance of with respect to .

As is the case in the biological world, there is a difference between genotype and phenotype. In general, there can be multiple representations (genotypes) that correspond to the same function (phenotype). That is, for some , with , still for all . However, this need not be the case. The goal then, is to find a representation that closely matches the phenotype of the ideal function, and the spirit of the local search is to allow only small changes in the genotype. Let the neighborhood of a representation be the set of possible mutations of .