精英学习的多维动态自适应人工鱼群算法
发布时间：2019-06-25 19:02
【摘要】：针对人工鱼群算法存在易陷入局部最优、鲁棒性差以及寻优精度低的问题,提出了精英学习的多维动态自适应人工鱼群算法.传统人工鱼群用欧式距离度量视野、步长,无法体现不同维度上鱼群的搜索进度.提出的算法为每个维度设定独立的视野和步长,从而定义了视野向量、步长矩阵及多维邻域,以此改进了鱼群的4种基本行为,使人工鱼个体能够根据鱼群分布情况自适应调整寻优范围.为了增加鱼群的全局性,降低人工鱼陷入局部最优的可能性,提出了一种人工鱼精英学习策略.仿真实验结果表明,该算法能有效地提高人工鱼群的寻优精度、寻优质量及鲁棒性,且提高了人工鱼群的全局搜索能力.
[Abstract]:In order to solve the problem that artificial fish swarm algorithm is easy to fall into local optimization, poor robustness and low optimization accuracy, a multi-dimensional dynamic adaptive artificial fish swarm algorithm based on elite learning is proposed. Traditional artificial fish stocks use Euclidean distance to measure visual field and step size, which can not reflect the search progress of fish stocks in different dimensions. The proposed algorithm sets an independent field of vision and step size for each dimension, thus defining the field vector, step size matrix and multi-dimensional neighborhood, which improves the four basic behaviors of fish stocks and enables artificial fish individuals to adaptively adjust the optimization range according to the distribution of fish stocks. In order to increase the global nature of fish stocks and reduce the possibility of artificial fish falling into local optimization, an elite learning strategy for artificial fish is proposed. The simulation results show that the algorithm can effectively improve the optimization accuracy, optimization quality and robustness of artificial fish stocks, and improve the global search ability of artificial fish stocks.
【作者单位】： 江南大学物联网工程学院;
【基金】：国家“八六三”高技术研究发展计划项目(2014AA041505)资助 国家自然科学基金项目(61572238)资助
【分类号】：TP18
，
本文编号：2505925
[Abstract]:In order to solve the problem that artificial fish swarm algorithm is easy to fall into local optimization, poor robustness and low optimization accuracy, a multi-dimensional dynamic adaptive artificial fish swarm algorithm based on elite learning is proposed. Traditional artificial fish stocks use Euclidean distance to measure visual field and step size, which can not reflect the search progress of fish stocks in different dimensions. The proposed algorithm sets an independent field of vision and step size for each dimension, thus defining the field vector, step size matrix and multi-dimensional neighborhood, which improves the four basic behaviors of fish stocks and enables artificial fish individuals to adaptively adjust the optimization range according to the distribution of fish stocks. In order to increase the global nature of fish stocks and reduce the possibility of artificial fish falling into local optimization, an elite learning strategy for artificial fish is proposed. The simulation results show that the algorithm can effectively improve the optimization accuracy, optimization quality and robustness of artificial fish stocks, and improve the global search ability of artificial fish stocks.
【作者单位】： 江南大学物联网工程学院;
【基金】：国家“八六三”高技术研究发展计划项目(2014AA041505)资助 国家自然科学基金项目(61572238)资助
【分类号】：TP18
，
本文编号：2505925
本文链接：https://www.wllwen.com/kejilunwen/zidonghuakongzhilunwen/2505925.html
