计算机科学 ›› 2017, Vol. 44 ›› Issue (3): 51-54.doi: 10.11896/j.issn.1002-137X.2017.03.013

• 2015全国高性能计算学术年会 • 上一篇    下一篇

基于MPI和OpenMP混合编程的非负矩阵分解并行算法

唐兵,Laurent BOBELIN,贺海武   

  1. 湖南科技大学计算机科学与工程学院 湘潭411201,中国科学院计算机网络信息中心 北京100190,中国科学院计算机网络信息中心 北京100190
  • 出版日期:2018-11-13 发布日期:2018-11-13
  • 基金资助:
    本文受中科院国际人才计划CAS PIFI(2016VTB028),湖南省自然科学基金(2015JJ3071)资助

Parallel Algorithm of Nonnegative Matrix Factorization Based on Hybrid MPI and OpenMP Programming Model

TANG Bing, Laurent BOBELIN and HE Hai-wu   

  • Online:2018-11-13 Published:2018-11-13

摘要: 非负矩阵分解(NMF)作为一种数据降维和特征提取的有效工具,已经在文本聚类、推荐系统等多个领域得到应用,但是其计算过程比较复杂。对此,提出一种基于MPI+OpenMP的混合层次化并行NMF方法,其充分利用基于MPI的消息传递模型和基于OpenMP的共享存储模型各自的优势,并基于多核节点集群进行测试。实验结果表明,所设计的并行NMF算法达到了较高的加速比,能有效处理高阶矩阵的非负分解,极大地提高了计算的效率。

关键词: 非负矩阵分解,并行算法,MPI,OpenMPI,可扩展

Abstract: Nonnegative matrix factorization (NMF) has been introduced as an efficient way to reduce the complexity of data and extracting character,and it has also been applied to various fields,such as recommendations and text clustering.However,the computation process of NMF is quite complex.In order to solve this problem,a hybrid parallel hierar-chical NMF algorithm based on OpenMP and MPI was presented in this paper,which makes full use of the advantages of both MPI-based message passing model and OpenMP-based shared storage model.The new algorithm is evaluated in a multi-core cluster environment,and experimental results demonstrate that it can achieve a high speed-up,and can be used to deal with large-scale NMF with a high efficiency.

Key words: Nonnegative matrix factorization,Parallel algorithm,MPI,OpenMPI,Scalability

[1] LEE D,SEUNG H.Learning the Parts of Objects by Nonnegative Matrix Factorization[J].Nature,1999,401:788-791.
[2] KOREN Y,BELL R,VOLINSKY C.Matrix Factorization Tech-niques for Recommender Systems[J].IEEE Computer,2009,42(8):40-49.
[3] CHEN Y,REGE M,DONG M,et al.Non-negative matrix factorization for semi-supervised data clustering[J].Knowledge and Information Systems,2008,17(3):355-379.
[4] CHOO J,LEE C,REDDY C,et al.UTOPIAN:User-Driven TopicModeling Based on Interactive Nonnegative Matrix Factorization[J].IEEE Transactions on Visualization and Computer Graphi-cs,2013,19(12):1992-2001.
[5] WANG W.Instantaneous Versus Convolutive Non-NegativeMatrix Factorization:Models,Algorithms and Applications to Audio Pattern Separation[M]∥Machine Audition:Principles,Algorithms and Systems.IGI Global,2010:353-370.
[6] LIAO R,ZHANG Y,GUAN J,et al.CloudNMF:A MapReduce Implementation of Nonnegative Matrix Factorization for Large-scale Biological Datasets[J].Genomics,Proteomics & Bioinformatics,2014,12(1):48-51.
[7] REN X X,TANG L,LI R F,et al.Study and Implementation of OpenMP Multi-thread Load Balance Scheduling Scheme[J].Computer Science,2010(11):148-151.(in Chinese) 任小西,唐玲,李仁发,等.OpenMP多线程负载均衡调度策略研究与实现[J].计算机科学,2010(11):148-151.
[8] FENG Y,ZHOU S Q.Research on development of mixed mode MPI+OpenMP applications[J].Computer Systems & Applications,2006,15(2):86-89.(in Chinese) 冯云,周淑秋.MPI+OpenMP 混合并行编程模型应用研究[J].计算机系统应用,2006,15(2):86-89.
[9] PAN W,CHEN L Y,ZHANG J H,et al.Research on MPI+OpenMP hybrid programming paradigm based on SMP cluster[J].Application Research of Computers,2009,26(12):4592-4594.(in Chinese) 潘卫,陈燎原,张锦华,等.基于SMP集群的MPI+OpenMP混合编程模型研究[J].计算机应用研究,2009,26(12):4592-4594.

No related articles found!
Viewed
Full text


Abstract

Cited

  Shared   
  Discussed   
No Suggested Reading articles found!