MIRSVM: Multi-instance support vector machine with bag representatives

作者：

Highlights：

• Novel bag-level representative multi-instance learning SVM framework is proposed.

• Primal and dual L1-SVM formulations and KKT conditions are devised and derived.

• Unique positive and negative bag-representative selector method is designed.

• The formulations use bag-level information to find an optimal hyperplane among bags.

• Results indicate the better performance of bag-level classifiers over other methods.

摘要

•Novel bag-level representative multi-instance learning SVM framework is proposed.•Primal and dual L1-SVM formulations and KKT conditions are devised and derived.•Unique positive and negative bag-representative selector method is designed.•The formulations use bag-level information to find an optimal hyperplane among bags.•Results indicate the better performance of bag-level classifiers over other methods.

论文关键词：Machine learning,Multiple-instance learning,Support vector machines,Bag-level multi-instance classification,Bag-representative selection

论文评审过程：Received 22 July 2017, Revised 16 January 2018, Accepted 10 February 2018, Available online 13 February 2018, Version of Record 23 February 2018.

论文官网地址：https://doi.org/10.1016/j.patcog.2018.02.007