A multiple-function toy model of exaptation in a genetic algorithm

In this paper we present a simple genetic algorithm consisting of a number of small niches, each with a different fitness function. The niches share a common genetic encoding and genotype-phenotype mapping, allowing for inter- niche migration of individuals. A notion of viability is introduced whereby population initialization produces viable individuals in one niche and is extremely unlikely to do so in all other niches. The niche fitness functions have been devised so as to demonstrate the gradual evolution of a population via multiple exaptation events where migrants seed, at each step, a new niche, adapt, and then spread to another in a predictable sequence. Such exaptation events take advantage of "hidden" relationships between fitness functions and allow evolving populations to explore regions of phenotype space that are otherwise inaccessible.