Infected genes evolutionary algorithm

Abstract

In this paper, we present a new Evolutionary Algorithm, called Infected Genes Evolutionary Algorithm. The idea behind this new algorithm is to focus the search process on certain genes, the 'Infected' genes. The result is a hard local search, but without diversity loss. Infection measures are associated with genes, reflecting a gene's quality in the context of a chromosome. These values evolve in each generation according to observations of the gene and the chromosome it belongs to. Search operators may then choose to search only the Infected genes, depending on an Infection Test. Results on two game-playing examples are shown where this new strategy outperformed the standard Genetic Algorithm. Besides the embedded hard local search property, this algorithm allows the implementation of crossover and mutation operators in a self-adaptive fashion, regarding the operator application probabilities.