{"status":"ok","message-type":"work","message-version":"1.0.0","message":{"indexed":{"date-parts":[[2026,4,2]],"date-time":"2026-04-02T01:46:34Z","timestamp":1775094394331,"version":"3.50.1"},"reference-count":48,"publisher":"Association for Computing Machinery (ACM)","issue":"3","license":[{"start":{"date-parts":[[2025,4,9]],"date-time":"2025-04-09T00:00:00Z","timestamp":1744156800000},"content-version":"vor","delay-in-days":0,"URL":"https:\/\/www.acm.org\/publications\/policies\/copyright_policy#Background"}],"content-domain":{"domain":["dl.acm.org"],"crossmark-restriction":true},"short-container-title":["ACM Trans. Knowl. Discov. Data"],"published-print":{"date-parts":[[2025,4,30]]},"abstract":"<jats:p>Logit perturbation refers to adding perturbation on logit, which has been shown to be capable of enhancing the robustness and generalization capabilities of deep neural networks in machine learning. However, studies on logit perturbation for multi-label learning are limited and they only consider the issue of class imbalance in the training data. Furthermore, the logit perturbation vectors in these methods are identical for negative classes containing different subclasses when multi-label learning is viewed as a multiple binary classification problem. This study investigates logit perturbation by exploring the characteristics of subclass-wise multi-label training data. First, the influence of the characteristics of multi-label training data on classification performance is analyzed in terms of the three data characteristics, namely, proportion, variance, and co-occurrence for each category (or subclass). Quantitative analyses reveal that variance differences among the subclasses in the negative class of a decomposed binary task also negatively impact the training performance, and if multiple characteristics affect simultaneously, the performance deterioration will be more severe. Second, theoretical analysis is performed for subclass-wise logit perturbation and a new subclass-wise logit perturbation method is proposed for multi-label learning. In our method, each class\/subclass has a carefully designed perturbation implementation according to its proportion, variance, and co-occurrence. Finally, our proposed method is further explained through a regularization view. Extensive experiments demonstrate that our method consistently enhances the generalization performance of popular depth networks on multi-label benchmark datasets.<\/jats:p>","DOI":"10.1145\/3715919","type":"journal-article","created":{"date-parts":[[2025,1,31]],"date-time":"2025-01-31T16:24:36Z","timestamp":1738340676000},"page":"1-40","update-policy":"https:\/\/doi.org\/10.1145\/crossmark-policy","source":"Crossref","is-referenced-by-count":3,"title":["Subclass-wise Logit Perturbation for Multi-label Learning"],"prefix":"10.1145","volume":"19","author":[{"ORCID":"https:\/\/orcid.org\/0000-0003-0693-435X","authenticated-orcid":false,"given":"Yu","family":"Zhu","sequence":"first","affiliation":[{"name":"Tianjin University, Tianjin, China and Tianjin University of Commerce, Tianjin, China"}],"role":[{"role":"author","vocabulary":"crossref"}]},{"ORCID":"https:\/\/orcid.org\/0000-0002-6106-6914","authenticated-orcid":false,"given":"Ou","family":"Wu","sequence":"additional","affiliation":[{"name":"Tianjin University, Tianjin, China and Hangzhou Institute for Advanced Study, University of Chinese Academy of Sciences, Hangzhou, China"}],"role":[{"role":"author","vocabulary":"crossref"}]},{"ORCID":"https:\/\/orcid.org\/0000-0002-3824-8306","authenticated-orcid":false,"given":"Fengguang","family":"Su","sequence":"additional","affiliation":[{"name":"Tianjin University, Tianjin, China"}],"role":[{"role":"author","vocabulary":"crossref"}]}],"member":"320","published-online":{"date-parts":[[2025,4,9]]},"reference":[{"key":"e_1_3_2_2_2","doi-asserted-by":"crossref","unstructured":"Alhuzali H. and Ananiadou S. 2021. SpanEmo: Casting multi-label emotion classification as span-prediction. arXiv:2101. 10038. Retrieved from https:\/\/doi:10.18653\/v1\/2021.eacl-main.135","DOI":"10.18653\/v1\/2021.eacl-main.135"},{"key":"e_1_3_2_3_2","doi-asserted-by":"publisher","DOI":"10.1109\/ICIP42928.2021.9506389"},{"key":"e_1_3_2_4_2","doi-asserted-by":"publisher","DOI":"10.1109\/ICCV.2019.00061"},{"key":"e_1_3_2_5_2","doi-asserted-by":"publisher","DOI":"10.1109\/CVPR.2019.00532"},{"key":"e_1_3_2_6_2","doi-asserted-by":"publisher","DOI":"10.1007\/978-3-030-58526-6_41"},{"key":"e_1_3_2_7_2","doi-asserted-by":"publisher","DOI":"10.1109\/CVPR.2019.00949"},{"key":"e_1_3_2_8_2","doi-asserted-by":"publisher","DOI":"10.1109\/TAFFC.2020.3034215"},{"key":"e_1_3_2_9_2","doi-asserted-by":"publisher","DOI":"10.1109\/ICCV.2017.205"},{"key":"e_1_3_2_10_2","doi-asserted-by":"crossref","unstructured":"Everingham M. Van Gool L. Williams C. K. Winn J. and Zisserman A. 2010. The pascal visual object classes (voc) challenge. International Journal of Computer Vision 88 (2010) 303\u2013338.","DOI":"10.1007\/s11263-009-0275-4"},{"key":"e_1_3_2_11_2","doi-asserted-by":"publisher","DOI":"10.1109\/CVPR46437.2021.01484"},{"key":"e_1_3_2_12_2","doi-asserted-by":"publisher","DOI":"10.1109\/CVPR.2016.90"},{"key":"e_1_3_2_13_2","doi-asserted-by":"publisher","unstructured":"Huang Y. Giledereli B. K\u00f6ksal A. \u00d6zg\u00fcr A. and Ozkirimli E. 2021. Balancing methods for multi-label text classification with long-tailed class distribution. arXiv:2109.04712. Retrieved from DOI: 10.18653\/v1\/2021.emnlp-main.643","DOI":"10.18653\/v1\/2021.emnlp-main.643"},{"key":"e_1_3_2_14_2","doi-asserted-by":"publisher","DOI":"10.1609\/aaai.v35i9.16974"},{"key":"e_1_3_2_15_2","doi-asserted-by":"publisher","DOI":"10.1609\/aaai.v34i05.6311"},{"key":"e_1_3_2_16_2","unstructured":"Kang B. Xie S. Rohrbach M. Yan Z. Gordo A. Feng J. and Kalantidis Y. 2019. Decoupling representation and classifier for long-tailed recognition. arXiv:1910.09217. Retrieved from https:\/\/arxiv.org\/abs\/1910.09217"},{"key":"e_1_3_2_17_2","doi-asserted-by":"publisher","DOI":"10.1609\/aaai.v36i2.20024"},{"key":"e_1_3_2_18_2","first-page":"13926","article-title":"Class-level logit perturbation","author":"Li M.","year":"2023","unstructured":"Li M., Su F., Wu O., and Zhang J. 2023. Class-level logit perturbation. IEEE Transactions on Neural Networks and Learning Systems, 13926\u201313940.","journal-title":"IEEE Transactions on Neural Networks and Learning Systems"},{"key":"e_1_3_2_19_2","doi-asserted-by":"publisher","DOI":"10.1109\/CVPR.2016.325"},{"key":"e_1_3_2_20_2","doi-asserted-by":"publisher","DOI":"10.1109\/CVPR46437.2021.00517"},{"key":"e_1_3_2_21_2","first-page":"1","article-title":"Multi-label image classification with a probabilistic label enhancement model","author":"Li X.","year":"2014","unstructured":"Li X., Zhao F., and Guo Y. 2014. Multi-label image classification with a probabilistic label enhancement model. In Conference on Uncertainty in Artificial Intelligence, 1\u201310.","journal-title":"Conference on Uncertainty in Artificial Intelligence"},{"key":"e_1_3_2_22_2","doi-asserted-by":"publisher","DOI":"10.1109\/ICCV.2017.324"},{"key":"e_1_3_2_23_2","doi-asserted-by":"publisher","DOI":"10.1007\/978-3-319-10602-1"},{"key":"e_1_3_2_24_2","doi-asserted-by":"publisher","DOI":"10.1145\/3451392"},{"key":"e_1_3_2_25_2","unstructured":"Reed S. Lee H. Anguelov D. Szegedy C. Erhan D. and Rabinovich A. 2015. Training deep neural networks on noisy labels with bootstrapping. arXiv:1412.6596. Retrieved from https:\/\/arxiv.org\/abs\/1412.6596"},{"key":"e_1_3_2_26_2","unstructured":"Menon A. K. Jayasumana S. Rawat A. S. Jain H. Veit A. and Kumar S. 2020. Long-tail learning via logit adjustment. arXiv:2007.07314. Retrieved from https:\/\/arxiv.org\/abs\/2007.07314"},{"key":"e_1_3_2_27_2","doi-asserted-by":"publisher","DOI":"10.1609\/aaai.v35i10.17098"},{"key":"e_1_3_2_28_2","doi-asserted-by":"publisher","DOI":"10.1109\/ICCV48922.2021.00015"},{"key":"e_1_3_2_29_2","doi-asserted-by":"publisher","DOI":"10.1007\/978-3-319-46478-7_29"},{"key":"e_1_3_2_30_2","doi-asserted-by":"publisher","DOI":"10.1109\/PRAI59366.2023.10332128"},{"key":"e_1_3_2_31_2","doi-asserted-by":"publisher","DOI":"10.1109\/CVPR.2016.308"},{"key":"e_1_3_2_32_2","doi-asserted-by":"publisher","DOI":"10.1109\/TNNLS.2022.3213522"},{"key":"e_1_3_2_33_2","doi-asserted-by":"crossref","first-page":"1","DOI":"10.1145\/3643853","article-title":"On the value of head labels in multi-label text classification","author":"Wang H.","year":"2024","unstructured":"Wang H., Peng C., Dong H., Feng L., Liu W., Hu T., and Chen G. 2024. On the value of head labels in multi-label text classification. ACM Transactions on Knowledge Discovery from Data, 1\u201321.","journal-title":"ACM Transactions on Knowledge Discovery from Data"},{"key":"e_1_3_2_34_2","doi-asserted-by":"publisher","DOI":"10.1109\/CVPR.2016.251"},{"key":"e_1_3_2_35_2","doi-asserted-by":"publisher","DOI":"10.1109\/TPAMI.2021.3052951"},{"key":"e_1_3_2_36_2","first-page":"12635","article-title":"Implicit semantic data augmentation for deep networks","volume":"32","author":"Wang Y.","year":"2019","unstructured":"Wang Y., Pan X., Song S., Zhang H., Huang G., and Wu C. 2019. Implicit semantic data augmentation for deep networks. Advances in Neural Information Processing Systems, 32, 12635\u201312644.","journal-title":"Advances in Neural Information Processing Systems"},{"key":"e_1_3_2_37_2","doi-asserted-by":"publisher","DOI":"10.1007\/978-3-030-58548-8_10"},{"key":"e_1_3_2_38_2","doi-asserted-by":"publisher","DOI":"10.1609\/aaai.v35i12.17244"},{"key":"e_1_3_2_39_2","doi-asserted-by":"publisher","DOI":"10.1109\/CVPR42600.2020.00090"},{"key":"e_1_3_2_40_2","doi-asserted-by":"publisher","DOI":"10.1007\/978-3-030-58589-1_39"},{"key":"e_1_3_2_41_2","doi-asserted-by":"publisher","DOI":"10.1109\/TNNLS.2021.3094304"},{"key":"e_1_3_2_42_2","first-page":"5820","article-title":"Attentionxml: Label tree-based attention-aware deep model for high-performance extreme multi-label text classification","author":"You R.","year":"2019","unstructured":"You R., Zhang Z., Wang Z., Dai S., Mamitsuka H., and Zhu S. 2019. Attentionxml: Label tree-based attention-aware deep model for high-performance extreme multi-label text classification. Advances in Neural Information Processing Systems, 5820\u20135830.","journal-title":"Advances in Neural Information Processing Systems"},{"key":"e_1_3_2_43_2","doi-asserted-by":"publisher","DOI":"10.1109\/CVPR.2017.219"},{"key":"e_1_3_2_44_2","first-page":"7472","volume-title":"International Conference on Machine Learning","author":"Zhang H.","year":"2019","unstructured":"Zhang H., Yu Y., Jiao J., Xing E., El Ghaoui L., and Jordan M. 2019. Theoretically principled trade-off between robustness and accuracy. In International Conference on Machine Learning. PMLR, 7472\u20137482."},{"key":"e_1_3_2_45_2","first-page":"11492","volume-title":"InInternational Conference on Machine Learning","author":"Xu H.","year":"2021","unstructured":"Xu H., Liu X., Li Y., Jain A., and Tang J. 2021. To be robust or to be fair: Towards fairness in adversarial training. In International Conference on Machine Learning. PMLR, 11492\u201311501. PMLR."},{"key":"e_1_3_2_46_2","doi-asserted-by":"crossref","unstructured":"Zhou D. W. Sun H. L. Ning J. Ye H. J. and Zhan D. C. 2024. Continual learning with pre-trained models: A survey. arXiv:2401.16386. Retrieved from https:\/\/arxiv.org\/abs\/2401.16386","DOI":"10.24963\/ijcai.2024\/924"},{"key":"e_1_3_2_47_2","first-page":"2978","article-title":"Multi-label continual learning using augmented graph convolutional network","author":"Du K.","year":"2023","unstructured":"Du K., Lyu F., Li L., Hu F., Feng W., Xu F., Xi X., and Cheng H. 2023. Multi-label continual learning using augmented graph convolutional network. IEEE Transactions on Multimedia, 2978\u20132992.","journal-title":"IEEE Transactions on Multimedia"},{"key":"e_1_3_2_48_2","first-page":"1952","volume-title":"International Conference on Machine Learning","author":"Chrysakis A.","year":"2020","unstructured":"Chrysakis A. and Moens M. F. 2020. Online continual learning from imbalanced data. In International Conference on Machine Learning, 1952\u20131961."},{"issue":"11","key":"e_1_3_2_49_2","first-page":"2579","article-title":"Visualizing data using t-SNE","volume":"9","author":"Van der Maaten L.","year":"2008","unstructured":"Van der Maaten L. and Hinton G. 2008. Visualizing data using t-SNE. Journal of Machine Learning Research, 9 (11), 2579\u20132605.","journal-title":"Journal of Machine Learning Research"}],"container-title":["ACM Transactions on Knowledge Discovery from Data"],"original-title":[],"language":"en","link":[{"URL":"https:\/\/dl.acm.org\/doi\/10.1145\/3715919","content-type":"unspecified","content-version":"vor","intended-application":"text-mining"},{"URL":"https:\/\/dl.acm.org\/doi\/pdf\/10.1145\/3715919","content-type":"unspecified","content-version":"vor","intended-application":"similarity-checking"}],"deposited":{"date-parts":[[2025,6,19]],"date-time":"2025-06-19T01:18:48Z","timestamp":1750295928000},"score":1,"resource":{"primary":{"URL":"https:\/\/dl.acm.org\/doi\/10.1145\/3715919"}},"subtitle":[],"short-title":[],"issued":{"date-parts":[[2025,4,9]]},"references-count":48,"journal-issue":{"issue":"3","published-print":{"date-parts":[[2025,4,30]]}},"alternative-id":["10.1145\/3715919"],"URL":"https:\/\/doi.org\/10.1145\/3715919","relation":{},"ISSN":["1556-4681","1556-472X"],"issn-type":[{"value":"1556-4681","type":"print"},{"value":"1556-472X","type":"electronic"}],"subject":[],"published":{"date-parts":[[2025,4,9]]},"assertion":[{"value":"2024-03-01","order":0,"name":"received","label":"Received","group":{"name":"publication_history","label":"Publication History"}},{"value":"2025-01-14","order":2,"name":"accepted","label":"Accepted","group":{"name":"publication_history","label":"Publication History"}},{"value":"2025-04-09","order":3,"name":"published","label":"Published","group":{"name":"publication_history","label":"Publication History"}}]}}