A Novel Genetic Algorithm Based on Immunity and Its Application

Fenggiang Zhao,Guangqiang Li,Jialu Du,Chen Guo,Hongying Hu,Ajith Abraham

DOI: https://doi.org/10.1109/his.2012.6421396

2012-01-01

Abstract:In this paper, a novel genetic algorithm based on immunity (GABI) on the basis of parallel genetic algorithms (PGA) is proposed in order to overcome some defects of them, such as premature and slow convergence rate. The global performance of the algorithm is improved by introducing immunity theory into PGA. This is revealed in the following two aspects. One is that the immune selection based on proposed adjustable geometric-progression rank-based selection can prevent the algorithm from premature. The other is that convergence rate can be accelerate by individual migration strategy between subpopulations based on immune memory mechanism. In this algorithm, the idea of multiple subpopulations evolution based on improved adaptive crossover and mutation is adopted. To be hybridized with the Powell method can further improve local searching performance of the algorithm. An example of layout design shows that GABI is feasible and effective.