Comparison of Genetic Operators on a General Genetic Algorithm Package

Abstract

Genetic Algorithm models the natural selection and evolution process and has been successful in areas where the traditional methods fall short. Most of the existing GA packages offer limited choices of encodings and genetic operators. To handle the variety of optimization problems, a more comprehensive GA package is designed and tested in this thesis. This GA package gives users a rich set of options on the encodings and genetic operators. The package is tested through two groups of experiments. The first group tests the general robustness using a variety of optimization problems. The second group compares the performances of different genetic operators by varying the encoding, selection, or crossover methods on certain input functions. The experiments results indicate that the package converged to satisfactory solutions. Based on the experiments conducted, tips regarding the performance of some genetic operators are provided.