Accelerating genetic algorithms: protected chromosomes and parallel processing

This paper proposes a new method for speeding up genetic algorithms (GAs) by temporarily protecting some components of chromosomes from ordinary genetic operators such as crossover and mutation. Protecting some components of chromosomes can reduce the search space, speeding up GAs that use this method. This new method is implemented on a subpopulation-type parallelised genetic algorithm. Each subpopulation protects some promising components of its chromosomes, and improves those components in cooperation with or competition against other subpopulations. Common components of chromesomes in a subpopulation are chosen as a protected part in that subpopulation, for convenience. Several travelling salesman problems are used to illustrate the accelerating effect of this method. The simulation results show that a GA using this method is able to search about twice as fast as a non-protecting parallelised GA