Genetic Algorithms comprise search and optimization strategies which are inspired by natural evolution: "survival of the fittest". In most of the best known basic genetic algorithms a binary coding of solution candidates is used. However for the DNA-coding, mother nature uses 4 purine bases: adenine (A), cytosine (C), guanine (G) and thymine (T). Following this idea, the present paper studies quaternary coded genetic algorithms and, based on high complex test functions, shows that these algorithms have a performance as good as their corresponding binary versions for problems with low dimensions and reach very fast an acceptable good fitness, if not the best, for high dimensional problems. For the first time it is shown, that the performance of genetic algorithms under a Gray code is sensitive to permutation of the columns of the code.

No abstract