| 基于监督的距离度量学习方法研究 |
| |
| 引用本文： | 战扬,金英,杨丰. 基于监督的距离度量学习方法研究[J]. 黑龙江电子技术, 2011, 0(12): 21-23 |
| |
| 作者姓名： | 战扬 金英 杨丰 |
| |
| 作者单位： | 黑龙江大学计算机科学技术学院,哈尔滨150080 |
| |
| 基金项目： | 黑龙江省教育厅科学技术研究项目（11511381）;黑龙江大学高层次人才（创新团队）支持计划（Hdtd2010-07）;黑龙江大学青年科学基金项目（QL201028） |
| |
| 摘 要： | 很多机器学习算法（比如K近邻算法），学习的效果非常依赖于输入数据的距离度量，距离度量学习的主要目标是通过训练样本学习出一个能够更有效反映样本空间的距离函数，在此距离函数下，同类样本具有较近的距离，异类样本具有较远的距离。对近年来基于监督的距离度量学习方法的基本思想和算法进行了研究，并对当前距离度量学习的热点进行了介绍。
|
| 关 键 词： | 距离度量学习 机器学习 K近邻分类器 |
Research on supervised distance metric learning |
| |
| ZHAN Yang,JIN Ying,YANG Feng. Research on supervised distance metric learning[J]. , 2011, 0(12): 21-23 |
| |
| Authors: | ZHAN Yang JIN Ying YANG Feng |
| |
| Affiliation: | ( School of Computer Science and Technology, Heilongjiang University, Harbin 150080, China) |
| |
| Abstract: | Lots of machine learning algorithms, such as K Nearest Neighbor, learning effect is very dependent on the distance measure of input data, the main objective of distance metric learning is through the training samples to learn a more effective distance function which can reflect the sample space. In this distance function, the same class samples have short distance, different class samples have far distance. This paper surveys the basic ideas and algorithms of supervised distance metric learning in recent years, and some current hot spots of distance metric learning are introduced. |
| |
| Keywords: | distance metric learning machine learning K nearest neighbor |
| 本文献已被 维普 等数据库收录！ |
|
