免疫记忆遗传算法及其完全收敛性研究

Research of Immune Memory Genetic Algorithm and its Complete Convergence

指出多模态优化中现有小生境遗传算法(NGA)和简单子群遗传算法(SSGA)无法实现完全收敛。受精英个体保留策略的启示,基于免疫系统的记忆细胞机制设计了免疫记忆遗传算法(IMGA),利用马尔柯夫链为数学工具,从理论上证明了NGA不能完全收敛而IMGA能够完全收敛。选择小生境遗传算法与该文算法进行了对比仿真实验,不仅验证了理论上的完全收敛性结论,同时验证了所提算法求解多模态问题的有效性、快速收敛能力及其收敛的稳定性。

By analyzing the mechanisms of Niche Genetic Algorithm(NGA) and Simple Sub-population Genetic Algorithm(SSGA) in now multi-modal optimization fields,the fault that they cannot convergent completely are pointed out and the concept of complete convergence is proposed.A new Immune Memory Genetic Algorithm(IMGA) based on mechanism of immune cells memory is designed.Using mathematical methods of Markov chains theory,it is proven that NGA is not completely convergent but IMGA is.The simulation experiments for NGA and IMGA are performed,and the results show that complete convergence proven above is right,also testify that IMGA has availability on solving multi-modal optimization problems,quickly convergence ability and wonderful stability of search results.