Bag dissimilarity regularized multi-instance learning

作者：

Highlights：

• We propose a bag dissimilarity regularized (BDR) framework for MIL.

• The BDR framework incorporates both implicit and explicit bag representations.

• We propose an explicit bag representation based on factor analysis and Fisher score.

• Two regular classifiers are transformed into BDR methods.

• The BDR methods outperform the comparison methods based on a single representation.

摘要

•We propose a bag dissimilarity regularized (BDR) framework for MIL.•The BDR framework incorporates both implicit and explicit bag representations.•We propose an explicit bag representation based on factor analysis and Fisher score.•Two regular classifiers are transformed into BDR methods.•The BDR methods outperform the comparison methods based on a single representation.

论文关键词：Multi-instance learning (MIL),Dissimilarity regularization,Fisher score

论文评审过程：Received 6 August 2021, Revised 17 December 2021, Accepted 8 February 2022, Available online 10 February 2022, Version of Record 16 February 2022.

论文官网地址：https://doi.org/10.1016/j.patcog.2022.108583