Zigzag mutation: a new mutation operator to improve the genetic algorithm

Sasan Harifi

doi:10.1007/s11042-023-15518-3

2023-05-01

PRE

Sasan Harifi *

Reza Mohamaddoust

DOI：10.1007/s11042-023-15518-3

原文链接

原文求助

摘要

En 中文

Genetic algorithm is an exploratory method inspired by Darwin's theory of natural evolution. This algorithm reflects the natural selection process in which suitable individuals are selected for reproduction to produce the offspring of the next generation. The genetic algorithm uses three main operators, namely selection, crossover, and mutation, each of which is involved in producing better strings or chromosomes. Among the three main operators, the mutation operator is one of the most important operators to achieve the optimal solution. The mutation operator is an intelligent mechanism for local search in the problem-solving search space. Mutations are therefore used to maintain population diversity and prevent premature convergence in the problem-solving process. In this study, to improve the genetic algorithm, a new type of mutation is introduced, which is called the Zigzag mutation. This mutation, by observance the zigzag pattern and making sudden and noticeable mutants in the gene compared to the existing mutations, make the local search in the problem space more efficient and helps to improve the genetic algorithm. In this paper, the proposed Zigzag mutation-based genetic algorithm is compared with six other genetic algorithms with different mutations in similar competitive conditions. The state-of-the-art mutations used in this study include Gaussian, Insertion, Inversion, Scramble, Swap, and Uniform, which are compared one by one with the proposed Zigzag mutation. In the experiments, 27 benchmark test functions are used to evaluate the performance. The evaluation results show superiority in 21 benchmark functions. According to the results, the presented method alone is better than other comparable methods in 77% of the benchmark functions. The share of the other six methods is only 23%. Also as an application, the presented improved genetic algorithm has been used in the segmentation of miscellaneous and aerial images. The results and statistical analysis show that the Zigzag mutation can improve the genetic algorithm and make it more efficient to solve the problems.

Keyword:

Genetic algorithm

Zigzag mutation

Gaussian mutation

Image segmentation

Benchmark test functions