文章摘要
唐 立,王利军.基于Spark的DA算法并行化研究[J].西昌学院学报(自然科学版),2019,33(4):66-69.
基于Spark的DA算法并行化研究
Study on DAAlgorithm Parallelization Based on Spark
  
DOI:
中文关键词: Spark  DA算法  并行化  大规模数据
英文关键词: Spark  DA algorithm  parallelization  mass data
基金项目:安徽省高校自然科学研究项目(KJ2019A0965);安徽省社会科学联合会课题(2018CX104)
作者单位
唐 立,王利军 安徽经济管理学院信息工程系合肥 230031 
摘要点击次数: 29
全文下载次数: 37
中文摘要:
      在对大规模数据进行蜻蜓算法优化时,由于要计算的维度过多,迭代次数过大,从而耗费大量运算时间,而基于Spark分 布式计算可以减少大数据运算的耗时。将DA算法在Spark分布式计算平台下进行并行计算,把蜻蜓种群被分配到各个节点, 每节点中蜻蜓个体信息通过多线程并行更新,然后共享全局最优解,从而提高大规模数据优化的运行速度。最后仿真实验的 验证是由4个测试函数进行测试,验证结果显示:在保证正确率的前提下,基于Spark的DA算法在对大规模数据优化的计算用 用时最少。
英文摘要:
      In the optimization of dragonfly algorithm for mass data, due to too many dimensions to be calculated and too many iterative computations, it takes a lot of operation time. However, the distributed algorithm based on Spark can reduce the operation time of big data. In this paper, the DA algorithm is used to conduct parallel computation on Spark distributed computing platform, and the dragonfly population is distributed to each node. The dragonfly individual information in each node is updated in parallel through multiple threads, and then the global optimal solution is shared, so as to improve the operation speed of mass data optimization. Finally, the verification of the simulation test is carried out through four test functions. The verification results show that on the premise of the accuracy ensured, the DA algorithm based on Spark takes the least time in the optimization calculation of mass data
查看全文   查看/发表评论  下载PDF阅读器
关闭