An implementation of new selection strategies in a genetic algorithm – population recombination and elitist refinement

Abstract

The present study aims to develop a new genetic algorithm utilizing population recombination and elitist refinement. Population recombination determines how a population and its elitist sub-population evolve. A whole population consists of three major sub-populations: the first is the current generation's elitist sub-population, the second is obtained from the pure crossover of the elitist sub-population with another existing sub-population, and the third is the result of a crossover between the elitist and random sub-populations. Genetic operations such as reproduction and crossover are applied among sub-populations during the process of population recombination. The refinement of the elitist sub-population is then implemented in order to improve the converged solution that was obtained from the recombination. The refinement of elitist sub-populations facilitates the locations of a more enhanced design by altering the binary values in chromosomes. The proposed method is verified through a number of nonlinear and/or multi-modal functions and constrained optimization problems.

Keywords:

• distributed genetic algorithm
• population recombination
• elitist sub-population
• elitist refinement
• tail evolution and fitting