Niche Search : an Evolutionary Algorithm

In this paper we describe niche search, a genetic-based opti-misation approach which is characterised by an evolutionary search on two layers: the individual layer (which is comparable to search described in other genetic algorithms), and the niche layer. Neither of these searches is directed: both individuals and niches evolve based on the selection of the ttest. The numerical results obtained by niche search are quite promising, as our implementation has successfully handled all the tests carried out. The computational performance is considerably better than that of other algorithms of the same family analysed in the literature. 1 Overview Niche search is a genetic algorithm version for whose conception special attention has been paid to the aspects that follow. First, encode the reproduction operations that occur in nature, at the cellular and molecular level, more precisely and completely than the implementation proposed in the canonical genetic algorithm (CGA). In particular, it is proposed a simulation of the meiosis process in the reproduction of our artiicial organisms. Meiosis is the main source of diversity in nature (for sexual reproduction), and is absent in the CGA. Second, the overall population of individuals | the ecosystem | is assembled into diierent groups, each of them occupying a particular niche. This parallels what happens in nature: individuals of a given species are in general not found neither alone nor assembled in the same global pool. Furthermore, the parameters that control the reproduction (mutation probability, crossover probability, etc.) may vary from place to place, as may vary the way in which diierent species exploit diierent parameters. In the approach proposed, each niche is supplied its own set of these parameters by the ecosystem, in an arbitrary and endogenous way. Finally, selection and reproduction act not only at the level of individuals, but also at the level of groups of individuals (niches). Individuals must compete in order to survive among their similars; niches must also compete between them for resources in the environment. Competition and selection between individuals of a particular niche is similar to what is proposed in the CGA; competition