The benefits of target relations: A comparison of multitask extensions and classifier chains

作者：

Highlights：

• Multi-objective approaches change the learning trajectory for the better.

• Using targets as additional inputs improves the generalization power of the classifiers.

• The order of adding targets is important in chaining.

• Exploiting target relations augments the learning process.

摘要

•Multi-objective approaches change the learning trajectory for the better.•Using targets as additional inputs improves the generalization power of the classifiers.•The order of adding targets is important in chaining.•Exploiting target relations augments the learning process.

论文关键词：Multitask learning,Multi-objective trees,Stacking,Classifier chains,Ensemble learning

论文评审过程：Received 10 May 2019, Revised 30 May 2020, Accepted 17 June 2020, Available online 24 June 2020, Version of Record 6 July 2020.

论文官网地址：https://doi.org/10.1016/j.patcog.2020.107507