| 基于委员会投票选择方法的主动学习的研究 |
| |
| 引用本文： | 赵悦,穆志纯. 基于委员会投票选择方法的主动学习的研究[J]. 太原理工大学学报, 2006, 37(4): 469-472 |
| |
| 作者姓名： | 赵悦 穆志纯 |
| |
| 作者单位： | 北京科技大学,信息工程学院,北京,100083 |
| |
| 摘 要： | 讨论了利用QBC(委员会投票选择)的主动学习方法来学习贝叶斯网络分类器,通过对基于投票熵和基于KL-divergence的QBC算法的研究,指出了两者存在的缺陷;同时提出了基于投票熵与类条件后验最大熵相结合的QBC算法。实验证明,该算法提高了分类器的精度,并使用了较少的训练样本,这有效解决了大数据量训练样本获得困难的情况下分类器建模问题。
|
| 关 键 词： | 主动学习 委员会投票选择 投票熵 最大熵 |
| 文章编号： | 1007-9432（2006）04-0469-04 |
| 收稿时间： | 2005-09-02 |
| 修稿时间： | 2005-09-02 |
Research on Query-by-Committee Method of Active Learning |
| |
| ZHAO Yue,MU Zhi-chun. Research on Query-by-Committee Method of Active Learning[J]. Journal of Taiyuan University of Technology, 2006, 37(4): 469-472 |
| |
| Authors: | ZHAO Yue MU Zhi-chun |
| |
| Affiliation: | School of Information Engineering ,Beijing University of Science and Technology, Beijing 100083, China |
| |
| Abstract: | This paper discussed the Query-by-Committee(QBC) methods of active learning.The disadvantages of QBC based on vote entropy and KL-divergence were presented.A new QBC method is presented by combining vote entropy and class conditional posterior maximum entropy for learning TAN classifier.Experimental results show that the new QBC algorithm improves the classification accuracy and uses less labeled training examples,which efficiently solves the problems on modeling the classifier under few labeled training examples. |
| |
| Keywords: | active learning Query-by-Committee vote entropy maximum entropy |
| 本文献已被 CNKI 维普 万方数据 等数据库收录！ |
|
