Multi-label metric transfer learning jointly considering instance space and label space distribution divergence
Multi-label learning deals with problems in which each instance is associated with a set of labels. Most multi-label learning algorithms ignore the potential distribution differences between the training domain and the test domain in the instance space and label space, as well as the intrinsic geome...
| Main Authors: | , , , , , , |
|---|---|
| Other Authors: | |
| Format: | Journal Article |
| Language: | English |
| Published: |
2019
|
| Subjects: | |
| Online Access: | https://hdl.handle.net/10356/86081 http://hdl.handle.net/10220/48323 |
| _version_ | 1811677759861161984 |
|---|---|
| author | Jiang, Siyu Xu, Yonghui Wang, Tengyun Yang, Haizhi Qiu, Shaojian Yu, Han Song, Hengjie |
| author2 | School of Computer Science and Engineering |
| author_facet | School of Computer Science and Engineering Jiang, Siyu Xu, Yonghui Wang, Tengyun Yang, Haizhi Qiu, Shaojian Yu, Han Song, Hengjie |
| author_sort | Jiang, Siyu |
| collection | NTU |
| description | Multi-label learning deals with problems in which each instance is associated with a set of labels. Most multi-label learning algorithms ignore the potential distribution differences between the training domain and the test domain in the instance space and label space, as well as the intrinsic geometric information of the label space. These restrictive assumptions limit the ability of the existing multi-label learning algorithms to classify between domains. To solve this problem, in this paper, we propose a novel distribution-adaptation-based method, the multi-label metric transfer learning (MLMTL), to relax these two assumptions and handle more general multi-label learning tasks effectively. In particular, MLMTL extends the maximum mean discrepancy method into multi-label classification by learning and adjusting the weights for the multi-labeled training instances. In this way, MLMTL bridges the instance distribution and label distribution divergence between training and test datasets. In addition, based on the balanced multi-label training data, we explore the intrinsic geometric information of the label space by encoding it into a distance metric learning framework. Extensive experiments on five benchmark datasets show that the proposed approach significantly outperforms the state-of-the-art multi-label learning algorithms. |
| first_indexed | 2024-10-01T02:42:29Z |
| format | Journal Article |
| id | ntu-10356/86081 |
| institution | Nanyang Technological University |
| language | English |
| last_indexed | 2024-10-01T02:42:29Z |
| publishDate | 2019 |
| record_format | dspace |
| spelling | ntu-10356/860812020-03-07T11:48:52Z Multi-label metric transfer learning jointly considering instance space and label space distribution divergence Jiang, Siyu Xu, Yonghui Wang, Tengyun Yang, Haizhi Qiu, Shaojian Yu, Han Song, Hengjie School of Computer Science and Engineering Transfer Learning Metric Learning DRNTU::Engineering::Computer science and engineering Multi-label learning deals with problems in which each instance is associated with a set of labels. Most multi-label learning algorithms ignore the potential distribution differences between the training domain and the test domain in the instance space and label space, as well as the intrinsic geometric information of the label space. These restrictive assumptions limit the ability of the existing multi-label learning algorithms to classify between domains. To solve this problem, in this paper, we propose a novel distribution-adaptation-based method, the multi-label metric transfer learning (MLMTL), to relax these two assumptions and handle more general multi-label learning tasks effectively. In particular, MLMTL extends the maximum mean discrepancy method into multi-label classification by learning and adjusting the weights for the multi-labeled training instances. In this way, MLMTL bridges the instance distribution and label distribution divergence between training and test datasets. In addition, based on the balanced multi-label training data, we explore the intrinsic geometric information of the label space by encoding it into a distance metric learning framework. Extensive experiments on five benchmark datasets show that the proposed approach significantly outperforms the state-of-the-art multi-label learning algorithms. Published version 2019-05-22T08:24:59Z 2019-12-06T16:15:37Z 2019-05-22T08:24:59Z 2019-12-06T16:15:37Z 2019 Journal Article Jiang, S., Xu, Y., Wang, T., Yang, H., Qiu, S., Yu, H., & Song, H. (2019). Multi-label metric transfer learning jointly considering instance space and label space distribution divergence. IEEE Access, 7, 10362-10373. doi:10.1109/ACCESS.2018.2889572 https://hdl.handle.net/10356/86081 http://hdl.handle.net/10220/48323 10.1109/ACCESS.2018.2889572 en IEEE Access © 2019 IEEE (Open Access). Translations and content mining are permitted for academic research only. Personal use is also permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information. 12 p. application/pdf |
| spellingShingle | Transfer Learning Metric Learning DRNTU::Engineering::Computer science and engineering Jiang, Siyu Xu, Yonghui Wang, Tengyun Yang, Haizhi Qiu, Shaojian Yu, Han Song, Hengjie Multi-label metric transfer learning jointly considering instance space and label space distribution divergence |
| title | Multi-label metric transfer learning jointly considering instance space and label space distribution divergence |
| title_full | Multi-label metric transfer learning jointly considering instance space and label space distribution divergence |
| title_fullStr | Multi-label metric transfer learning jointly considering instance space and label space distribution divergence |
| title_full_unstemmed | Multi-label metric transfer learning jointly considering instance space and label space distribution divergence |
| title_short | Multi-label metric transfer learning jointly considering instance space and label space distribution divergence |
| title_sort | multi label metric transfer learning jointly considering instance space and label space distribution divergence |
| topic | Transfer Learning Metric Learning DRNTU::Engineering::Computer science and engineering |
| url | https://hdl.handle.net/10356/86081 http://hdl.handle.net/10220/48323 |
| work_keys_str_mv | AT jiangsiyu multilabelmetrictransferlearningjointlyconsideringinstancespaceandlabelspacedistributiondivergence AT xuyonghui multilabelmetrictransferlearningjointlyconsideringinstancespaceandlabelspacedistributiondivergence AT wangtengyun multilabelmetrictransferlearningjointlyconsideringinstancespaceandlabelspacedistributiondivergence AT yanghaizhi multilabelmetrictransferlearningjointlyconsideringinstancespaceandlabelspacedistributiondivergence AT qiushaojian multilabelmetrictransferlearningjointlyconsideringinstancespaceandlabelspacedistributiondivergence AT yuhan multilabelmetrictransferlearningjointlyconsideringinstancespaceandlabelspacedistributiondivergence AT songhengjie multilabelmetrictransferlearningjointlyconsideringinstancespaceandlabelspacedistributiondivergence |