论文部分内容阅读
针对传统进化算法在计算效能方面存在的一些问题,借鉴协同进化算法的思想,提出了一种融合免疫机制的协同进化模型。该模型通过多个子种群各自分别进化以保持整个种群的多样性。在每次迭代进化过程中,各个子种群分别选择精英抗体并进行免疫记忆。随后各个子种群分别以不同的算法进行变异。若变异后抗体的适应度降低,则利用精英抗体对其进行引导操作。群体间的协作包括子种群间若干个抗体的随机交叉和子种群间的大规模迁移。最终进行免疫代谢,去除群中的弱适应度个体。算法反复迭代进行以上操作,直至达到既定目标或预定的循环迭代次数。通过对13个标准测试函数进行的仿真实验显示,该模型在搜索最优解或满意解时均优于传统的进化算法,同时在寻优效率上有较大的提升。
Aiming at some problems of traditional evolutionary algorithms in computational efficiency and using the idea of the co-evolutionary algorithm, a co-evolutionary model of fusion mechanism is proposed. The model evolves separately from each other to maintain the diversity of the entire population. During each iteration evolution, each subpopulation chooses elite antibodies separately and carries out immune memory. Subsequent subpopulations vary by different algorithms. If the variation of antibodies to reduce the fitness, the use of elite antibodies to guide its operation. Collaboration among populations includes random crossover of several antibodies between subpopulations and large-scale migration between subpopulations. The final immune metabolism, removal of the group of weak fitness individuals. The algorithm iterates over and over until the set target or predetermined number of iterations of the loop are reached. The simulation experiments on 13 standard test functions show that the proposed model outperforms the traditional evolutionary algorithms when searching for the optimal solution or the satisfactory solution, and at the same time has a great improvement on the optimization efficiency.