技术问题请联系 网站管理员
|
报告信息 |
| 题目：全局高秩（和局部低秩）的缺失多标记学习 |
| 报告人：陈松灿 教授，南京航空航天大学 |
| 摘要：Multi-label learning (MLL) is an important machine learning paradigm where completing the missed labels is a key. One of popular MLL routes works under the assumption of global low-rankness in label completion. However, we find it does often not hold even in those commonly-used benchmark datasets. In this talk, we will conversely illustrate the global high-rankness in single/Multiview/Contrastive MLLs is essential, then propose concise yet effective learning approaches. |