基于反向学习的人工蜂群算法收敛性研究

Research on the Convergence Theory of Artificial Bee Colony Algorithm Based on Opposition Leaning

目的研究人工蜂群算法在搜索方面表现较好而在开采方面表现相对薄弱及人工蜂群算法收敛性证明较少的情况,解决标准的人工蜂群算法容易除局部最优和早熟收敛问题.方法提出基于反向学习的人工蜂群算法,使算法跳出局部最优及早熟收敛,更利于找到最优解;利用Markov链等理论对基于反向学习的人工蜂群算法进行简单的收敛性分析,给出算法的基本实现步骤,并通过一组测试函数进行实验.结果采用GABC算法对种群大小为80,最大循环数为5000,独立运行30次进行实验,实验数据表明,D=30的OABC的实验数据略差于GABC,而D=60的OABC的标准差数据好于GABC和ABC算法.实验结果表明改进后的算法在许多方面比标准的人工蜂群算法有更好的表现.结论收敛性分析表明基于反向学习的人工蜂群算法具有较好的收敛性.

The purpose of this paper is to solve that artificial bee colony （ABC）has shown to be competitive with some conventional biological-inspired algorithms, however, conventional ABC is good at exploration but poor at exploitation and it has little convergence demonstration;At the same time, in order to solve the standard artificial bee colony algorithm, the problem of local opti- mum and premature convergence is easy to be solved. So it presents an opposite artificial Bee Col- ony Algorithm which is contribute to finding the optimum solution and give a simple convergence demonstration according to Markov. The GABC algorithm is used for population size is 80, the maximum number of cycles was 5000,30 of the trials independently, the experimental data show that the experimental data of D = 30 OABC is slightly less than GABC ,D = 60 and the standard de- viation of OABC data is better than thnt of GABC and ABC algorithm. The experimental conse-quences tested on a series of numeric benchmark functions show that OABC algorithm can outper- form ABC algorithm in most of the experiments and it also has excellent performance in conver- gences.