Real-Valued Genetic Algorithms with Disagreements

This paper introduces a new mutation operator for real-valued genetic algorithms that refines the evolutionary process using disagreements. After a short introduction, we describe the new concept theoretically and then we exemplify it by defining a Gaussian distribution-based disagreements operator: the 6 sigma-GAD. We transform two common real-valued genetic algorithms into their disagreements-enabled counterparts and we conduct several tests proving that our newly obtained algorithms perform better because they gain strengthened neighborhood focus using partial disagreements and enhanced exploration capabilities through extreme disagreements.