计算机与现代化 ›› 2012, Vol. 198 ›› Issue (2): 176-179.doi: 10.3969/j.issn.1006-2475.2012.02.047

• 应用与开发 • 上一篇    下一篇

基于OpenMP的动态模式识别并行计算

孙珊   

  1. 华南理工大学自动化科学与工程学院,广东广州510641
  • 收稿日期:2011-08-30 修回日期:1900-01-01 出版日期:2012-02-24 发布日期:2012-02-24

Parallel Computing of Dynamic Pattern Recognition Based on OpenMP

SUN Shan   

  1. School of Automation Science and Engineering, South China University of Technology, Guangzhou 510641, China
  • Received:2011-08-30 Revised:1900-01-01 Online:2012-02-24 Published:2012-02-24

摘要: 考虑到实现确定学习理论中的动态模式识别过程耗时过多,提出一种适用于识别过程计算的多核并行技术。以压气机Mansoux模型为研究背景,首先对其模式获取和识别的过程进行简单描述,其次,在四核PC的硬件平台上,使用OpenMP编程,实现了对动态模式识别的并行计算。通过设置不同的线程数,讨论了几个影响并行程序性能的重要因素。结果表明,要综合考虑各种因素的影响才能设计出高效的并行识别程序。

关键词: 并行计算, OpenMP, 确定学习, 模式识别

Abstract: Given that the realization of dynamic pattern recognition of deterministic learning theory is timeconsuming, a multicore parallel computation technique suitable for recognition process is proposed. Firstly, based on the classical Mansoux compressor model, the acquirement of patterns and the process of recognition are described in brief. Secondly, parallel computing is applied to dynamic pattern recognition by using OpenMP programming on four cores PC hardware platform. Factors which influence the performance of program are discussed by setting different thread numbers. The result shows that the only way to design parallel recognition program with high performance is taking kinds of elements into account.

Key words: parallel computing, OpenMP, deterministic learning, pattern recognition

中图分类号: