Class-Level Logit Perturbation
Article
Li, Mengyang, Su, Fengguang, Wu, Ou and Zhang, Ji. 2023. "Class-Level Logit Perturbation." IEEE Transactions on Neural Networks and Learning Systems. https://doi.org/10.1109/TNNLS.2023.3273355
Article Title | Class-Level Logit Perturbation |
---|---|
ERA Journal ID | 4458 |
Article Category | Article |
Authors | Li, Mengyang, Su, Fengguang, Wu, Ou and Zhang, Ji |
Journal Title | IEEE Transactions on Neural Networks and Learning Systems |
Number of Pages | 15 |
Year | 2023 |
Publisher | IEEE (Institute of Electrical and Electronics Engineers) |
Place of Publication | United States |
ISSN | 1045-9227 |
1941-0093 | |
2162-237X | |
2162-2388 | |
Digital Object Identifier (DOI) | https://doi.org/10.1109/TNNLS.2023.3273355 |
Web Address (URL) | https://ieeexplore.ieee.org/document/10130785 |
Abstract | Features, logits, and labels are the three primary data when a sample passes through a deep neural network (DNN). Feature perturbation and label perturbation receive increasing attention in recent years. They have been proven to be useful in various deep learning approaches. For example, (adversarial) feature perturbation can improve the robustness or even generalization capability of learned models. However, limited studies have explicitly explored for the perturbation of logit vectors. This work discusses several existing methods related to class-level logit perturbation. A unified viewpoint between regular/irregular data augmentation and loss variations incurred by logit perturbation is established. A theoretical analysis is provided to illuminate why class-level logit perturbation is useful. Accordingly, new methodologies are proposed to explicitly learn to perturb logits for both the single-label and multilabel classification tasks. Meta-learning is also leveraged to determine the regular or irregular augmentation for each class. Extensive experiments on benchmark image classification datasets and their long-tail versions indicated the competitive performance of our learning method. As it only perturbs on logit, it can be used as a plug-in to fuse with any existing classification algorithms. All the codes are available at https://github.com/limengyang1992/lpl. |
Keywords | Adversarial training |
ANZSRC Field of Research 2020 | 460299. Artificial intelligence not elsewhere classified |
Public Notes | The accessible file is the accepted version of the paper. Please refer to the URL for the published version. |
Byline Affiliations | Tianjin University, China |
University of Southern Queensland |
Permalink -
https://research.usq.edu.au/item/z262y/class-level-logit-perturbation
Download files
31
total views17
total downloads0
views this month0
downloads this month