Evolution through cooperation: The Symbiotic Algorithm

Abstract

This article describes a new problem solving paradigm: the Symbiotic Algorithm. Problem solutions are considered as embedded organisms whose genetic materials are expressed as evaluable phenotypes. This organic hierarchy can be viewed as a biosphere containing the whole set of creatures manipulated by the algorithm. Being immortal, organisms do not replicate. The only changes occuring in the biosphere are the creation or destruction of symbiotic relationships between organisms. To let these relationships appear and evolve, we provide a fitness function which decides the outcome of organic interplay. The higher an organism's fitness is, the more likely it is to remain unchanged and the more likely it is to invade weaker participants.

Some experiments, addressing the optimization of sinusoidal functions, show that the organic hierarchy adopts configurations in which appear substructures corresponding to optimal solutions. Moreover, the use of multimodal fitness functions induce phenotype distributions matching the fitness function peaks.