TCS Journal 2006 Journal Article
A genetic system based on simulated crossover of sequences of two-bit genes
- Marco Carpentieri
We introduce a genetic model based on simulated crossover of fixed sequences of two-bit genes. Results are (1) a lower bound on population size is exhibited such that a transition takes the stochastic finite population genetic system near the next state of the deterministic infinite population genetic system (provided both begin in the same state); (2) states and dynamics of the deterministic infinite population genetic system are derived for arbitrary (finite) fitness functions (expressed in terms of multivariate polynomials); (3) in the case of quadratic fitness defined by weight matrices with m nonnull entries it is shown that each state transition can be implemented in time O ( m + l ), where l is the chromosome length; (4) the genetic algorithm (implementing the proposed infinite population system) is experimentally compared with the infinite population genetic algorithm with bit-based simulated crossover for the max-cut problem; the results show that the extension to sequences of genes with four alleles is useful to improve performances.