Competitive multi-species evolution has been useful in optimization applications due to its better results by comparison to a single problem solver population. Its coevolution with a problem creator population pushes both populations to increasingly better solutions, a phenomenon called arms-race.
On the other hand, symbiosis is a kind of cooperative coevolution, which has been gaining relevance in biology. Artificial symbiogenetic coevolution has been shown to improve evolutionary optimization algorithms by a specialization of the different components of the symbiotic collaboration. In this case of cooperative coevolution there is a kind of division of labor between the different types of symbionts. Each host is combined with a parasite forming a collaboration. A collaboration is evaluated as a solution to the optimization problem. This is repeated for different hosts and parasites. These two types of populations are evaluated by averaging the fitness of each collaboration they were involved in, and each population is submitted to an independent artificial evolutionary process. Artificial symbiogenetic evolution has been proving useful in solving deceptive problems – a class of functions that is specially difficult to optimize due to the fact that the optimum is surrounded by regions of low quality solutions.
Artificial evolutionary models are inspired by nature, but when used as engineering tools do not need to maintain a strict correspondence with their natural counterparts. The main goal of engineering is to obtain efficient tools, in this case designed to solve optimization problems. Taking this into account, we further explore different approaches of evolutionary algorithms and their operators, that one may consider unrealistic by comparison to nature.
We end by suggesting that artificial symbiogenetic evolution may also be interesting to model specific aspects of natural evolution – the artificial models allowing exploration of configurations that may be hard to build in vivo.