Population Sizing Scheme for Genetic Algorithm

Cell placement is a phase in the chip design process, in which cells representing well-defined functions are assigned physical locations. Cell placement is an NP-complete problem, for which we intend to devise an adaptive genetic algorithm. Genetic algorithms have many parameters such as population size, mutation rate, crossover rate, and selection strategy, which are constants most of the time and need to be carefully set for efficient implementation. However, adaptive approaches tend to vary one or more of those parameters as the process evolve. In this work, we propose a scheme to adjust the population size in a way that provides a balance between exploration and exploitation, hence result in a time-efficient implementation of genetic algorithms. We compare this scheme with three sizing schemes proposed in the literature.