Raising GA performance by simultaneous tuning of selective pressure and recombination disruptiveness

In many Genetic Algorithms applications the objective is to find a (near-)optimal solution using a limited amount of computation. Given these requirements it is difficult to find a good balance between exploration and exploitation. Usually such a balance is found by tuning the various parameters (like the selective pressure, population size, the mutation- and crossover rate) of the Genetic Algorithm. As an alternative we propose simultaneous tuning of the selective pressure and the disruptiveness of the recombination operators. Our experiments show that the combination of a proper selective pressure and a highly disruptive recombination operator yields superior performance. The reduction mechanism used in a Steady-State GA has a strong influence on the optimal crossover disruptiveness. Using the worst fitness deletion strategy the building blocks present in the current best individuals are always preserved. This releases the crossover operator from the burden to maintain good building blocks and allows us to tune crossover disruptiveness to improve the search for better individuals.