Diversified dictionaries for multi-instance learning

作者：

Highlights：

• 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.

摘要

Highlights•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.

论文关键词：Multi-instance learning,Diversified learning,Dictionary learning

论文评审过程：Received 29 May 2016, Revised 28 July 2016, Accepted 22 August 2016, Available online 26 August 2016, Version of Record 24 December 2016.

论文官网地址：https://doi.org/10.1016/j.patcog.2016.08.026