摘要
为满足大规模文本快速分类的需求,在传统文本分类方案基础上,利用GPU强大的并行吞吐量,提出了一种大规模并行文本分类方案。为验证该方案的有效性,在多个平台上进行充分的实验分析。结果表明,该方案比传统的分类方案具有10倍以上的加速比。
To satisfy the need for fast classification of large scale texts, a new solution of parallel text classification is introduced, which is based on classical text classification solution and utilizes the powerful throughput of GPU. Extensive lab experiments are done in diff＆ent platforms to verify the effectiveness of the solution. The result shows that it has 10X speedup compared with classical solution.
出处
《计算机工程与应用》
CSCD
2012年第8期141-143,206,共4页
Computer Engineering and Applications
基金
国家自然科学基金(No.61103038)
关键词
图形处理器
统一计算设备架构
朴素贝叶斯
并行文本分类
graphical processing unit
compute unified device architecture
native Bayes
parallel text categorization
