| 多分类Boosting算法的一致性 |
| |
| 引用本文： | 唐轶.多分类Boosting算法的一致性[J].计算机工程与应用,2006,42(18):27-28,39. |
| |
| 作者姓名： | 唐轶 |
| |
| 作者单位： | 湖北大学数学与计算机科学学院,武汉,430062 |
| |
| 摘 要： | 根据样本容量适当选取正则参数可以使得多分类Boosting算法具有一致性。通过分析正则参数对多分类Boosting算法推广能力的影响,建立了正则参数与算法一致性之间的联系。据此得到了Boosting算法具有一致性的充分条件。在样本集确定时,该条件可作为多分类Boosting算法选择正则参数的依据。
|
| 关 键 词： | Boosting算法 一致性 无限样本一致性 |
| 文章编号： | 1002-8331-（2006）18-0027-02 |
| 收稿时间： | 2006-01 |
| 修稿时间： | 2006-01 |
The Consistency of the Multi-Category Classification Boosting Algorithm |
| |
| Tang Yi.The Consistency of the Multi-Category Classification Boosting Algorithm[J].Computer Engineering and Applications,2006,42(18):27-28,39. |
| |
| Authors: | Tang Yi |
| |
| Affiliation: | Faculty of Mathematics and Computer Science,Hubei University,Wuhan 430062 |
| |
| Abstract: | The consistency of the multi-category classification Boosting algorithms can be obtained by choosing suitable regularization parameters according to the number of the samples.Considering the relationships between the regularization parameters and the generalibility of the boosting algorithms,we establish the sufficient conditions to insure the consistency of the Boosting algorithms.Moreover we achieve a method based on such conditions for selecting regularization parmeters of Boosting algorithms. |
| |
| Keywords: | Boosting algorithm consistency infinite-sample consistency |
| 本文献已被 CNKI 维普 万方数据 等数据库收录！ |
