1 Genetic Algorithms based Improved Sampling

Abstract: In this paper, we study extensions of Genetic Algorithm (GA) to incorporate improved sampling capacity. While existing nondeterministic algorithms including GA are feasible to implement and effective for specific domains, they however fail to solve a wider range of problems especially for the hard optimization problems. We have developed a fast-navigating GA (FNGA) using associated-memory (AM) based crossover operation. The AM stores the crossover-position based participating best chromosome’s subpart from the passing generations. The stored subpart gets more trials which helps navigate faster as well as to obtain accurate solution. However, this process increases the similarity within population and reduce directory faster. To mitigate it, we immediately trigger the robust-searching (termed Twin-Removal GA or, TRGA) mode which maintain diversity optimally by replacing closely similar chromosomes with optimal number of random chromosomes after FNGA’s role in every generation. The optimally diverge chromosomes generated by TRGA can introduce potential subpart to enhance the performance of FNGA further. Thus we combine FNGA and TRGA and named the combination KGA. The proposed extensions of GA were empirically studied and found to perform generic enough to solve a wide ranges of benchmark problems very effectively and outperformed state-of-the-art algorithms