Statistics-based adaptive non-uniform crossover for genetic algorithms

Through the population, genetic algorithm (GA) implicitly maintains the statistics about the search space. This implicit statistics can be used explicitly to enhance GA's performance. Inspired by this idea, a statistics-based adaptive non-uniform crossover, called SANUX, has been proposed. SANUX uses the statistics information of the alleles in each locus to adaptively calculate the swapping probability of that locus for crossover. A simple triangular function has been used to calculate the swapping probability. In this paper two different functions, the trapezoid and exponential functions, are investigated for SANUX instead of the triangular function. The experiment results show that both functions further improve the performance of SANUX across a typical set of GA's test problems.