On convergence and optimality of genetic algorithms
Witold Kosiński, Stefan Kotowski, Zbyszek Michalewicz · 2010
An action of genetic algorithm could be represented in the search space as a random Markovian process. The question concerning its asymptotic stability properties is discussed. Conditions under which genetic algorithm is convergent, are formulated. Then the existence of an operator to which infinite long iterations of the genetic algorithms tend, is shown. This operator describes optimal genetic algorithm in probabilistic sense.