Toward an understanding of the quality and efficiency of model building for genetic algorithms

Abstract

This paper investigates the linkage model building for genetic algorithms. By assuming a given quality of the linkage model, an analytical model of time to convergence is derived. Given the computational cost of building the linkage model, an estimated total computational time is obtained by using the derived time-to-convergence model. The models are empirically verified. The results can be potentially used to decide whether applying a linkage-identification technique is worthwhile and give a guideline to speed up the linkage model building. © Springer-Verlag Berlin Heidelberg 2004.