Diversified dictionaries for multi-instance learning

详细信息    查看全文

• 作者：Maoying Qiao^a ; ^{Maoying.Qiao@student.uts.edu.au}Author Vitae ; Liu Liu^a ; ^{Liu.Liu@student.uts.edu.au}Author Vitae ; Jun Yu^b ; ^{yujun@hdu.edu.cn}Author Vitae ; Chang Xu^c ; ^{changxu1989@gmail.com}Author Vitae ; Dacheng Tao^c ; ^{Dacheng.Tao@uts.edu.au}Author Vitae
• 关键词：Multi-instance learning ; Diversified learning ; Dictionary learning
• 刊名：Pattern Recognition
• 出版年：2017
• 出版时间：April 2017
• 年：2017
• 卷：64
• 期：Complete
• 页码：407-416
• 全文大小：527 K
• 卷排序：64

文摘

This paper presents a supervised diversified dictionaries MIL to address the problem of bridging instance-level representations to bag-level labels. The proposed method exploits bag-level label information for training class-specific dictionaries. The proposed method introduces a diversity regulariser into the class-specific dictionaries to avoid ambiguity between them. To the best of our knowledge, this is the first time that the diversity prior is introduced to solve the MIL problems.