| 并行支持向量机算法及其应用 |
| |
| 引用本文： | 林继鹏,刘君华,凌振宝.并行支持向量机算法及其应用[J].吉林大学学报(信息科学版),2004,22(5):453-457. |
| |
| 作者姓名： | 林继鹏 刘君华 凌振宝 |
| |
| 作者单位： | 西安交通大学,电气学院,陕西,西安,710049;吉林大学,电子科学与工程学院,吉林,长春,130026 |
| |
| 摘 要： | 针对当前支持向量机计算效率的不足,提出了改进的并行支持向量机技术.该算法能有效使内积运算、向量数据更新、矩阵向量实现并行计算,并且数据之间的通信时间能和向量更新时间重叠,从而提高了计算效率,并能保证泛化能力.数据仿真结果表明,与改进前的算法相比,2 500个样本下能节省时间30%左右,样本量增大时,效果更为明显.
|
| 关 键 词： | 统计学习 支持向量机 并行支持向量机 泛化能力 |
| 文章编号： | 1671-5896(2004)05-0453-05 |
| 修稿时间： | 2004年3月29日 |
Arithmetic based on PSVM and its applications |
| |
| LIN Ji-peng,LIU Jun-hua,LING Zhen-bao.Arithmetic based on PSVM and its applications[J].Journal of Jilin University:Information Sci Ed,2004,22(5):453-457. |
| |
| Authors: | LIN Ji-peng LIU Jun-hua LING Zhen-bao |
| |
| Institution: | LIN Ji-peng~1,LIU Jun-hua~1,LING Zhen-bao~2 |
| |
| Abstract: | A new method of PSVM(Parallel Support Vector Machine)derived from traditional TSVM(Tactic Support Vector Machine) was proposed to improve calculating efficiency. With this method,the inner products, vector updates and matrix-vector multiplication can be realized independently and the communication time was overlapped with vector updates without changing predictive ability. In data simulation, more than 30% time is saved in PSVM with 2500 samples,and much better in large scale sample sets. |
| |
| Keywords: | statistical learning parallel support vector machine support vector machine predictive ablity |
| 本文献已被 CNKI 维普 万方数据 等数据库收录！ |
|
