摘要
在海量数据背景下,传统的基于单个计算节点的算法很难满足分析要求。考察了一种分布式贝叶斯估计方法,通过在每台机器上单独运行蒙特卡洛抽样并做加权平均可以有效地解决算法效率问题。将该方法应用于基于广义极值模型的二元响应变量回归分析,并探讨其实用性。模拟研究表明分布式算法比传统方法更有效。
In the background of massive data,it is difficult to meet the analysis requirements for traditional one-node based algorithm.This paper considers a distributed Bayesian estimation method to solve the GEV based general linear regression model by running a separate Monte Carlo algorithm on each machine.The method is applied to regression analysis of binary response variables based on generalized extreme value model.The results show that the proposed distributed Bayesian regression algorithm is much faster than the traditional algorithm in the simulated data sets studying.
出处
《华东理工大学学报（自然科学版）》
CSCD
北大核心
2017年第4期559-562,共4页
Journal of East China University of Science and Technology
基金
国家高科技研究发展("863")计划(2015AA20107)
上海市经信委"软件和集成电路产业发展专项资金"(140304)
