| 基于增量学习的超球支持向量机设计 |
| |
| 引用本文： | 张曦煌,须文波.基于增量学习的超球支持向量机设计[J].计算机工程与应用,2006,42(13):66-68,76. |
| |
| 作者姓名： | 张曦煌 须文波 |
| |
| 作者单位： | 江南大学信息工程学院,江苏无锡,214122 |
| |
| 摘 要： | 增量学习是通过从已知样本出发对未知样本进行识别和分类,并能够继续学习的方法和原则。论文在分析了HS-SVM的理论基础后,基于Joachims的直推式SVM分类算法,提出了直推式THS-SVM算法,同时,独立提出了简单自学习的SHS-SVM学习方法。THS-SVM和SHS-SVM能够在训练过程中不断学习无标签样本的信息。实验表明将THS-SVM和SHS-SVM用于基于内容的图像检索是有效的。
|
| 关 键 词： | 超球 支持向量机 增量学习 直推式 |
| 文章编号： | 1002-8331-（2006）13-0066-03 |
| 收稿时间： | 2005-09 |
| 修稿时间： | 2005-09 |
The Design of Hyper-sphere SVM Based on Incremental Learning |
| |
| Zhang Xihuang,Xu Wenbo.The Design of Hyper-sphere SVM Based on Incremental Learning[J].Computer Engineering and Applications,2006,42(13):66-68,76. |
| |
| Authors: | Zhang Xihuang Xu Wenbo |
| |
| Affiliation: | School of Information Engineering, Southern Yangtse University, Wuxi, Jiangsu 214122 |
| |
| Abstract: | The SVM incremental learning is a method which can continue SVM learning through unlabeled samples classification using labeled samples.After analyzing the Hyper-sphere SVM theory,a transductive inference HS-SVM based on Joachims'TSVM and a incremental learning method called SHS-SVM are presented.The two algorithms can transfer unlabeled sampled information to Hyper-sphere SVM during training and are effective proven by CBIR experiments. |
| |
| Keywords: | hyper-sphere Support Vector Machine incremental leaming transductive inference |
| 本文献已被 CNKI 维普 万方数据 等数据库收录！ |
|
