| 支持向量机大规模快速算法 |
| |
| 引用本文： | 王越,夏书银,马胜. 支持向量机大规模快速算法[J]. 计算机工程与设计, 2012, 33(5): 1983-1987 |
| |
| 作者姓名： | 王越 夏书银 马胜 |
| |
| 作者单位： | 1. 重庆理工大学计算机科学与工程学院,重庆,400054
2. 长安工业有限责任公司研发中心,重庆,401120 |
| |
| 基金项目： | 重庆市科委攻关项目森林健康监测系统基金项目(CSTC,2009AC2068) |
| |
| 摘 要： | 在不影响泛化能力的情况下,针对现有的主要分块算法、大规模缩减策略和分解算法等内存占用较大、训练精度下降和收敛速度过慢等缺点,改进了现有的SMO算法,融合分块算法和分解算法,提出了最小序列分块算法(CSMO).仿真结果表明,该算法与libsvm等现有的典型的支持向量机算法相比,能够减小内存占用,并能以很高的精度接近全局最优解.
|
| 关 键 词： | 支持向量机 分块 分解 最小序列优化 全局最优解 |
Support vector machine large-scale fast algorithm |
| |
| WANG Yue , XIA Shu-yin , MA Sheng. Support vector machine large-scale fast algorithm[J]. Computer Engineering and Design, 2012, 33(5): 1983-1987 |
| |
| Authors: | WANG Yue XIA Shu-yin MA Sheng |
| |
| Affiliation: | 1.Institute of Computer Science and Engineering,Chongqing University of Technology,Chongqing 400054,China; 2.Research and Development Center,Chang Industrial Limited Liability Company,Chongqing 401120,China) |
| |
| Abstract: | Without affecting the generalization,to overcome the shortcomings of too large memory’s being used,training accuracy’s reduced,convergence’s slow down in the mainly existing chunking algorithms,large-scale samples’ curtail strategy,decomposition algorithms and so on.The SMO algorithm is improved,and combinated with the chunking and decomposition algorithm,chunking sequence minimization algorithm is proposed(CSMO).Simulation results show that,compared with other libsvm and other existing typical support vector machine algorithm,the algorithm can reduce the memory footprint are able to be close with the global optimal solution with a very high accuracy. |
| |
| Keywords: | SVM chunking decomposition SMO global optimal solution |
| 本文献已被 CNKI 万方数据 等数据库收录！ |
|
