图 书 馆
关于图上距离的注意事项A note of caution regarding distances on graphs |
|
| 课程网址: | http://videolectures.net/icml2010_von_luxburg_ancr/ |
| 主讲教师: | Ulrike von Luxburg |
| 开课单位: | 马克斯普朗克研究所 |
| 开课时间: | 2010-07-20 |
| 课程语种: | 英语 |
| 中文简介: |  |
| 课程简介: | Non-geometric data is often represented in form of a graph where edges represent similarity or local relationships between instances. One elegant way to exploit the global structure of the graph is implemented by the commute distance (also known as resistance distance). Supposedly it has the property that vertices from the same cluster are "close" to each other whereas vertices from different clusters are "far" from each other. We study the behavior of the commute distance as the size of the underlying graph increases. We prove that the commute distance converges to an expression that does not take into account the structure of the graph and that is completely meaningless as a distance function on the graph. Consequently, the use of the raw commute distance for machine learning purposes is strongly discouraged for large graphs and in high dimensions. We suspect that a similar behavior holds for several other distances on graphs. |
| 关 键 词: | 非几何数据; 图形; 局部关系 |
| 课程来源: | 视频讲座网 |
| 最后编审: | 2019-04-25：cwx |
| 阅读次数: | 26 |