Learning common and label-specific features for multi-Label classification with correlation information
作者：
Highlights：
• We extract common and label specific features simultaneously in learning process.
• We adopt a novel assumption to exploit label correlations, that is, if two.labels are strongly correlated, their corresponding output tend to be similar.
• In addition to label correlations, we also consider instance correlations by calculating instance similarity using K-Nearest Neighbor.
摘要
•We extract common and label specific features simultaneously in learning process.•We adopt a novel assumption to exploit label correlations, that is, if two.labels are strongly correlated, their corresponding output tend to be similar.•In addition to label correlations, we also consider instance correlations by calculating instance similarity using K-Nearest Neighbor.
论文关键词：Multi-label classification,Label-specific features,Common features,Instance correlation
论文评审过程：Received 27 July 2020, Revised 7 August 2021, Accepted 15 August 2021, Available online 16 August 2021, Version of Record 21 August 2021.
论文官网地址：https://doi.org/10.1016/j.patcog.2021.108259
