Domain adaption via feature selection on explicit feature map
In most domain adaption approaches, all features are used for domain adaption. However, often, not every feature is beneficial for domain adaption. In such cases, incorrectly involving all features might cause the performance to degrade. In other words, to make the model trained on the source domain...
| Main Authors: | , , , , , |
|---|---|
| Other Authors: | |
| Format: | Journal Article |
| Language: | English |
| Published: |
2020
|
| Subjects: | |
| Online Access: | https://hdl.handle.net/10356/140627 |
| _version_ | 1826112275173343232 |
|---|---|
| author | Deng, Wan-Yu Lendasse, Amaury Ong, Yew-Soon Tsang, Ivor Wai-Hung Chen, Lin Zheng, Qing-Hua |
| author2 | School of Computer Science and Engineering |
| author_facet | School of Computer Science and Engineering Deng, Wan-Yu Lendasse, Amaury Ong, Yew-Soon Tsang, Ivor Wai-Hung Chen, Lin Zheng, Qing-Hua |
| author_sort | Deng, Wan-Yu |
| collection | NTU |
| description | In most domain adaption approaches, all features are used for domain adaption. However, often, not every feature is beneficial for domain adaption. In such cases, incorrectly involving all features might cause the performance to degrade. In other words, to make the model trained on the source domain work well on the target domain, it is desirable to find invariant features for domain adaption rather than using all features. However, invariant features across domains may lie in a higher order space, instead of in the original feature space. Moreover, the discriminative ability of some invariant features such as shared background information is weak, and needs to be further filtered. Therefore, in this paper, we propose a novel domain adaption algorithm based on an explicit feature map and feature selection. The data are first represented by a kernel-induced explicit feature map, such that high-order invariant features can be revealed. Then, by minimizing the marginal distribution difference, conditional distribution difference, and the model error, the invariant discriminative features are effectively selected. This problem is NP-hard to be solved, and we propose to relax it and solve it by a cutting plane algorithm. Experimental results on six real-world benchmarks have demonstrated the effectiveness and efficiency of the proposed algorithm, which outperforms many state-of-the-art domain adaption approaches. |
| first_indexed | 2024-10-01T03:04:18Z |
| format | Journal Article |
| id | ntu-10356/140627 |
| institution | Nanyang Technological University |
| language | English |
| last_indexed | 2024-10-01T03:04:18Z |
| publishDate | 2020 |
| record_format | dspace |
| spelling | ntu-10356/1406272020-08-12T08:26:56Z Domain adaption via feature selection on explicit feature map Deng, Wan-Yu Lendasse, Amaury Ong, Yew-Soon Tsang, Ivor Wai-Hung Chen, Lin Zheng, Qing-Hua School of Computer Science and Engineering Engineering::Computer science and engineering Distribution Distance Domain Adaption In most domain adaption approaches, all features are used for domain adaption. However, often, not every feature is beneficial for domain adaption. In such cases, incorrectly involving all features might cause the performance to degrade. In other words, to make the model trained on the source domain work well on the target domain, it is desirable to find invariant features for domain adaption rather than using all features. However, invariant features across domains may lie in a higher order space, instead of in the original feature space. Moreover, the discriminative ability of some invariant features such as shared background information is weak, and needs to be further filtered. Therefore, in this paper, we propose a novel domain adaption algorithm based on an explicit feature map and feature selection. The data are first represented by a kernel-induced explicit feature map, such that high-order invariant features can be revealed. Then, by minimizing the marginal distribution difference, conditional distribution difference, and the model error, the invariant discriminative features are effectively selected. This problem is NP-hard to be solved, and we propose to relax it and solve it by a cutting plane algorithm. Experimental results on six real-world benchmarks have demonstrated the effectiveness and efficiency of the proposed algorithm, which outperforms many state-of-the-art domain adaption approaches. Accepted version This work was supported inpart by the National Science Foundation of China under Grant 61572399, Grant 61721002, Grant 61532015, Grant 61532004, and Grant 61472315, in part by the National Key Research and Development Program of China under Grant 2016YFB1000903, in part by the Shaanxi New Star of Science and Technology under Grant 2013KJXX-29, in part by the New Star Team of Xi’an University of Posts and Telecommunications, in part by the Provincial Key Disciplines Construction Fund of General Institutions of Higher Education in Shaanxi, in part by the Data Science and Artificial Intelligence Center at the Nanyang Technological University, in part by the ASTAR Thematic Strategic Research Program under Grant 1121720013, in part by the Computational Intelligence Research Laboratory at NTU, in part by the ARC Future Fellowship under Grant FT130100746, in part by the ARC Linkage Project under Grant LP150100671, and in part by the ARC Discovery Project under Grant DP180100106. 2020-06-01T02:42:15Z 2020-06-01T02:42:15Z 2018 Journal Article Deng, W.-Y., Lendasse, A., Ong, Y.-S., Tsang, I. W.-H., Chen, L., & Zheng, Q.-H. (2019). Domain adaption via feature selection on explicit feature map. IEEE Transactions on Neural Networks and Learning Systems, 30(4), 1180-1190. doi:10.1109/TNNLS.2018.2863240 2162-237X https://hdl.handle.net/10356/140627 10.1109/TNNLS.2018.2863240 30176608 2-s2.0-85052657350 4 30 1180 1190 en IEEE Transactions on Neural Networks and Learning Systems © 2018 IEEE. Personal use of this material is permitted. Permission from IEEE must be obtained for all other uses, in any current or future media, including reprinting/republishing this material for advertising or promotional purposes, creating new collective works, for resale or redistribution to servers or lists, or reuse of any copyrighted component of this work in other works. The published version is available at: https://doi.org/ 10.1109/TNNLS.2018.2863240 application/pdf |
| spellingShingle | Engineering::Computer science and engineering Distribution Distance Domain Adaption Deng, Wan-Yu Lendasse, Amaury Ong, Yew-Soon Tsang, Ivor Wai-Hung Chen, Lin Zheng, Qing-Hua Domain adaption via feature selection on explicit feature map |
| title | Domain adaption via feature selection on explicit feature map |
| title_full | Domain adaption via feature selection on explicit feature map |
| title_fullStr | Domain adaption via feature selection on explicit feature map |
| title_full_unstemmed | Domain adaption via feature selection on explicit feature map |
| title_short | Domain adaption via feature selection on explicit feature map |
| title_sort | domain adaption via feature selection on explicit feature map |
| topic | Engineering::Computer science and engineering Distribution Distance Domain Adaption |
| url | https://hdl.handle.net/10356/140627 |
| work_keys_str_mv | AT dengwanyu domainadaptionviafeatureselectiononexplicitfeaturemap AT lendasseamaury domainadaptionviafeatureselectiononexplicitfeaturemap AT ongyewsoon domainadaptionviafeatureselectiononexplicitfeaturemap AT tsangivorwaihung domainadaptionviafeatureselectiononexplicitfeaturemap AT chenlin domainadaptionviafeatureselectiononexplicitfeaturemap AT zhengqinghua domainadaptionviafeatureselectiononexplicitfeaturemap |