Discriminative feature generation for classification of imbalanced data
作者:
Highlights:
• A novel supervised discriminative feature generation (DFG) method using attention maps in the feature space is presented.
• We combine transfer learning and adversarial feature augmentation to complement their drawbacks.
• Extensive experiments on various datasets show that the proposed method enhances the augmentation of label-preserved and diverse features, and the classification results are significantly improved.
摘要
•A novel supervised discriminative feature generation (DFG) method using attention maps in the feature space is presented.•We combine transfer learning and adversarial feature augmentation to complement their drawbacks.•Extensive experiments on various datasets show that the proposed method enhances the augmentation of label-preserved and diverse features, and the classification results are significantly improved.
论文关键词:Imbalanced classification,Generative adversarial networks,Discriminative feature generation,Transfer learning,Feature map regularization
论文评审过程:Received 14 October 2020, Revised 28 January 2021, Accepted 2 September 2021, Available online 4 September 2021, Version of Record 24 September 2021.
论文官网地址:https://doi.org/10.1016/j.patcog.2021.108302