Reflections on the Geno- and the Phenotype A Phenotypic Approach to Cooperation for Genetic Algorithms

This article reports on the progress of implementing symbiotic networks for the Train Scheduling Problem. This particular paper discusses the use of GAs in symbiotic networks, which results in a framework that is comparable to the model of coevolving cooperative genetic algorithms (CCGA) proposed by Potter and de Jong. In CCGAs, distributed sub-populations of genetic algorithms, species try to find a global solution through cooperation. Such cooperation is normally established through a form of credit assignment. The problem of assigning such credits is often a matter of choice and experiment. The experiments conducted in this paper uses symbiotic networks as framework for cooperation. This approach could be considered phenotypic, as it creates optimising feedback loops between species through their mutual behaviour and a form of communication through stress signals. GAs can be used to generate such behaviour, while fitness can be easily mapped to the stress signals.