多线程并行构建三支概念

发布时间：2018-02-27 19:17

本文关键词： 形式概念分析三支概念分析形式背景多线程　出处：《西安交通大学学报》2017年03期 　论文类型：期刊论文

【摘要】：针对三支概念分析理论中三支概念数量庞大、构建耗时的问题,提出了一种三支概念的并行构建算法PCbO3C。PCbO3C以提高三支概念的构建效率为目标,在三支概念串行构建算法CbO3C的基础上进行并行化改进,利用多线程技术并行计算给定形式背景的所有核心三支概念。并行化处理借鉴了算法PCbO的思想,通过串行算法CbO3C计算出第L层的所有三支概念,并存放到P个队列中,第L层当前生成的三支概念循环依次放入P个队列中,以使算法达到较高的负载均衡;创建P个线程,利用CbO3C并行处理P个队列中的三支概念,使得CPU资源得到充分利用。由于多线程间没有同步操作,使得PCbO3C算法的整体效率得到了进一步提高。为了验证算法PCbO3C的效率,在8核CPU环境下对多组UCI和随机数据进行实验,实验结果表明:PCbO3C速度上明显优于CbO3C,当线程数不超过8时,线程数每增加1倍,并行算法的速度可以提高约67%。
[Abstract]:In order to solve the problem that the number of three concepts in the theory of three-branch conceptual analysis is huge and the construction time consuming, a parallel construction algorithm PCbO3C.PCbO3C is proposed to improve the efficiency of constructing three-branch concepts. On the basis of the serial construction algorithm CbO3C of three concepts, this paper improves parallelization, uses multithreading technology to calculate all the core three branch concepts of a given formal background in parallel. The parallel processing uses the idea of the algorithm PCbO for reference. All three concepts of L layer are calculated by serial algorithm CbO3C, and put into P queue side by side. The three concepts that are currently generated in L layer are put into P queue in turn, so that the algorithm achieves higher load balance, and P thread is created. Using CbO3C to process the three branches of P queue in parallel, the CPU resources are fully utilized. The overall efficiency of the PCbO3C algorithm is further improved because there is no synchronous operation between multithreads. In order to verify the efficiency of the algorithm PCbO3C, Experiments on multiple groups of UCI and random data are carried out under the environment of 8-core CPU. The experimental results show that the speed of the parallel algorithm is obviously better than that of CbO3C. When the number of threads is not more than 8:00, the speed of parallel algorithm can be increased by about 67 when the number of threads is less than 8:00 and the number of threads increases by 1 times.
【作者单位】：西安电子科技大学计算机学院;
【基金】：国家自然科学基金资助项目(11371014,11071281) 陕西省自然科学基础研究计划资助项目(2014JM8306)
【分类号】：TP301.6

【相似文献】