Improving GA performance by using Maximal Hyper-Rectangle analysis and relative fitness
MetadataShow full item record
In this thesis, we present two techniques to improve the performance of the genetic algorithm (GA). First we use Maximal Hyper-Rectangle (MHR) analysis to improve GA search reliability. We propose a method to find a sufficiently large MHR for new individual insertion in GA with polynomial computational complexity. Second, we propose an idea of relative fitness to improve increasing the convergence and searching more space. The individuals are selected for reproduction according to both of their global rank and regional rank. We apply the two techniques to some GA problems, and the results demonstrate their merits.