Coevolution has been used as optimization technique bothsuccessfully and unsuccessfully. Successful optimization showsintegration of information at the individual level over many fitnessevaluation events and over many generations. Alternative outcomes of theevolutionary process, e.g. red queen dynamics or speciation, preventsuch integration. Why coevolution leads to integration of information orto alternative evolutionary outcomes is generally unclear. We studycoevolutionary optimization of the density classification task incellular automata in a spatially explicit, two-species model. We findoptimization at the individual level, i.e. evolution of cellularautomata that are good density classifiers. However, when we globallymix the populations, which prevents the formation of spatial patterns,we find typical red queen dynamics in which cellular automata classifyall cases to a single density class regardless their actual density.Thus, we get different outcomes of the evolutionary process dependent ona small change in the model. We compare the two processes leading to thedifferent outcomes in terms of the diversity of the two populations atthe level of the genotype and at the level of the phenotype
展开▼