Effects of population size on the performance of genetic algorithms and the role of crossover

Authors:
Yu-An Zhang;Qinglian Ma;Makoto Sakamoto;Hiroshi Furutani
Affiliations:
Interdisciplinary Graduate School of Agriculture and Engineering, University of Miyazaki, Miyazaki, Japan;Faculty of Engineering, University of Miyazaki, Miyazaki, Japan 889-2192;Faculty of Engineering, University of Miyazaki, Miyazaki, Japan 889-2192;Faculty of Engineering, University of Miyazaki, Miyazaki, Japan 889-2192
Venue:
Artificial Life and Robotics
Year:
2010

Citing 4
Cited 0

On the Mean Convergence Time of Evolutionary Algorithms without Selection and Mutation

PPSN III Proceedings of the International Conference on Evolutionary Computation. The Third Conference on Parallel Problem Solving from Nature: Parallel Problem Solving from Nature
Effects of Population Size on Computational Performance of Genetic Algorithm on Multiplicative Landscape

ICNC '07 Proceedings of the Third International Conference on Natural Computation - Volume 03
A markov chain framework for the simple genetic algorithm

Evolutionary Computation
Schema analysis of average fitness in multiplicative landscape

GECCO'03 Proceedings of the 2003 international conference on Genetic and evolutionary computation: PartI

Quantified Score

Hi-index	0.00

Visualization

Abstract

One of the most important parameters in the application of genetic algorithms (GAs) is the population size N. In many cases, the choice of N determines the quality of the solutions obtained. The study of GAs with a finite population size requires a stochastic treatment of evolution. In this study, we examined the effects of genetic fluctuations on the performance of GA calculations. We considered the role of crossover by using the stochastic schema theory within the framework of the Wright-Fisher model of Markov chains. We also applied the diffusion approximation of the Wright-Fisher model. In numerical experiments, we studied effects of population size N and crossover rate pc on the success probability S. The success probability S is defined as the probability of obtaining the optimum solution within the limit of reaching the stationary state. We found that in a GA with pc, the diffusion equation can reproduce the success probability S. We also noted the role of crossover, which greatly increases S.