当前位置: X-MOL 学术Appl. Soft Comput. › 论文详情
Two-level label recovery-based label embedding for multi-label classification with missing labels
Applied Soft Computing ( IF 5.472 ) Pub Date : 2020-11-20 , DOI: 10.1016/j.asoc.2020.106868
Yibin Wang; Weijie Zheng; Yusheng Cheng; Dawei Zhao

Different from single-label learning, multi-label learning has rich semantic information. Label embedding obtains the inherent intelligence of the label space by projecting the label space into a latent one. However, the label space is inevitably incomplete. The missing label data will lead to the label embedding model capturing incomplete inherent information. Therefore, label data recovery becomes particularly important. However, the label correlations recovery mechanism only considers the label correlations but ignores the objective existence of instance-correlation information. The information we obtain from the perspective of label correlations alone will produce inadequate and incorrect phenomena when a large number of labels are lost or wrong. In this paper, a two-level label recovery mechanism is used not only to recover label data from the label perspective but also to take full advantage of the instance correlations in the label space. Although the two-level label recovery mechanism considers both instances and labels, it ignores the label space after recovery to capture its essential information. Therefore, combining the two-level semantic information, we propose a label embedding for multi-label classification based on a two-level label recovery mechanism about missing labels model for incomplete datasets, which is the two-level label recovery. Firstly, the two-level label space is projected to the inherent space, and then the instance and label correlations information are captured in the inherent space. Finally, it is applied to the label prediction of classifiers. The algorithm is trained and tested on several complete and incomplete multi-label datasets. It shows that the proposed algorithm has a good classification effect on the largely missed datasets. The results and statistical hypothesis tests further verify the effectiveness of the proposed algorithm.

更新日期:2020-11-21
全部期刊列表>>
ERIS期刊投稿
欢迎阅读创刊号
自然职场,为您触达千万科研人才
spring&清华大学出版社
城市可持续发展前沿研究专辑
Springer 纳米技术权威期刊征稿
全球视野覆盖
施普林格·自然新
chemistry
物理学研究前沿热点精选期刊推荐
自然职位线上招聘会
欢迎报名注册2020量子在线大会
化学领域亟待解决的问题
材料学研究精选新
GIANT
ACS ES&T Engineering
ACS ES&T Water
屿渡论文,编辑服务
ACS Publications填问卷
阿拉丁试剂right
苏州大学
林亮
南方科技大学
朱守非
胡少伟
有机所林亮
隐藏1h前已浏览文章
课题组网站
新版X-MOL期刊搜索和高级搜索功能介绍
ACS材料视界
上海纽约大学
浙江大学
廖矿标
天合科研
x-mol收录
试剂库存
down
wechat
bug