Long-Tailed Graph Representation Learning via Dual Cost-Sensitive Graph Convolutional Network
Deep learning algorithms have seen a massive rise in popularity for remote sensing over the past few years. Recently, studies on applying deep learning techniques to graph data in remote sensing (e.g., public transport networks) have been conducted. In graph node classification tasks, traditional gr...
| Main Authors: | , , , , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
MDPI AG
2022-07-01
|
| Series: | Remote Sensing |
| Subjects: | |
| Online Access: | https://www.mdpi.com/2072-4292/14/14/3295 |
| _version_ | 1827604424003944448 |
|---|---|
| author | Yijun Duan Xin Liu Adam Jatowt Hai-tao Yu Steven Lynden Kyoung-Sook Kim Akiyoshi Matono |
| author_facet | Yijun Duan Xin Liu Adam Jatowt Hai-tao Yu Steven Lynden Kyoung-Sook Kim Akiyoshi Matono |
| author_sort | Yijun Duan |
| collection | DOAJ |
| description | Deep learning algorithms have seen a massive rise in popularity for remote sensing over the past few years. Recently, studies on applying deep learning techniques to graph data in remote sensing (e.g., public transport networks) have been conducted. In graph node classification tasks, traditional graph neural network (GNN) models assume that different types of misclassifications have an equal loss and thus seek to maximize the posterior probability of the sample nodes under labeled classes. The graph data used in realistic scenarios tend to follow unbalanced long-tailed class distributions, where a few majority classes contain most of the vertices and the minority classes contain only a small number of nodes, making it difficult for the GNN to accurately predict the minority class samples owing to the classification tendency of the majority classes. In this paper, we propose a dual cost-sensitive graph convolutional network (DCSGCN) model. The DCSGCN is a two-tower model containing two subnetworks that compute the posterior probability and the misclassification cost. The model uses the cost as ”complementary information” in a prediction to correct the posterior probability under the perspective of minimal risk. Furthermore, we propose a new method for computing the node cost labels based on topological graph information and the node class distribution. The results of extensive experiments demonstrate that DCSGCN outperformed other competitive baselines on different real-world imbalanced long-tailed graphs. |
| first_indexed | 2024-03-09T05:59:31Z |
| format | Article |
| id | doaj.art-00b46418a6694fa1bdabedaa52f71e7e |
| institution | Directory Open Access Journal |
| issn | 2072-4292 |
| language | English |
| last_indexed | 2024-03-09T05:59:31Z |
| publishDate | 2022-07-01 |
| publisher | MDPI AG |
| record_format | Article |
| series | Remote Sensing |
| spelling | doaj.art-00b46418a6694fa1bdabedaa52f71e7e2023-12-03T12:10:25ZengMDPI AGRemote Sensing2072-42922022-07-011414329510.3390/rs14143295Long-Tailed Graph Representation Learning via Dual Cost-Sensitive Graph Convolutional NetworkYijun Duan0Xin Liu1Adam Jatowt2Hai-tao Yu3Steven Lynden4Kyoung-Sook Kim5Akiyoshi Matono6National Institute of Advanced Industrial Science and Technology Tokyo Waterfront, 2 Chome-3-26 Aomi, Koto City, Tokyo 135-0064, JapanNational Institute of Advanced Industrial Science and Technology Tokyo Waterfront, 2 Chome-3-26 Aomi, Koto City, Tokyo 135-0064, JapanDepartment of Computer Science, University of Innsbruck, Innrain 52, 6020 Innsbruck, AustriaFaculty of Library, Information and Media Science, University of Tsukuba, 1 Chome-1-1 Tennodai, Tsukuba 305-8577, JapanNational Institute of Advanced Industrial Science and Technology Tokyo Waterfront, 2 Chome-3-26 Aomi, Koto City, Tokyo 135-0064, JapanNational Institute of Advanced Industrial Science and Technology Tokyo Waterfront, 2 Chome-3-26 Aomi, Koto City, Tokyo 135-0064, JapanNational Institute of Advanced Industrial Science and Technology Tokyo Waterfront, 2 Chome-3-26 Aomi, Koto City, Tokyo 135-0064, JapanDeep learning algorithms have seen a massive rise in popularity for remote sensing over the past few years. Recently, studies on applying deep learning techniques to graph data in remote sensing (e.g., public transport networks) have been conducted. In graph node classification tasks, traditional graph neural network (GNN) models assume that different types of misclassifications have an equal loss and thus seek to maximize the posterior probability of the sample nodes under labeled classes. The graph data used in realistic scenarios tend to follow unbalanced long-tailed class distributions, where a few majority classes contain most of the vertices and the minority classes contain only a small number of nodes, making it difficult for the GNN to accurately predict the minority class samples owing to the classification tendency of the majority classes. In this paper, we propose a dual cost-sensitive graph convolutional network (DCSGCN) model. The DCSGCN is a two-tower model containing two subnetworks that compute the posterior probability and the misclassification cost. The model uses the cost as ”complementary information” in a prediction to correct the posterior probability under the perspective of minimal risk. Furthermore, we propose a new method for computing the node cost labels based on topological graph information and the node class distribution. The results of extensive experiments demonstrate that DCSGCN outperformed other competitive baselines on different real-world imbalanced long-tailed graphs.https://www.mdpi.com/2072-4292/14/14/3295graph convolutional networkimbalanced data classificationcost-sensitive learningsemi-supervised learning |
| spellingShingle | Yijun Duan Xin Liu Adam Jatowt Hai-tao Yu Steven Lynden Kyoung-Sook Kim Akiyoshi Matono Long-Tailed Graph Representation Learning via Dual Cost-Sensitive Graph Convolutional Network Remote Sensing graph convolutional network imbalanced data classification cost-sensitive learning semi-supervised learning |
| title | Long-Tailed Graph Representation Learning via Dual Cost-Sensitive Graph Convolutional Network |
| title_full | Long-Tailed Graph Representation Learning via Dual Cost-Sensitive Graph Convolutional Network |
| title_fullStr | Long-Tailed Graph Representation Learning via Dual Cost-Sensitive Graph Convolutional Network |
| title_full_unstemmed | Long-Tailed Graph Representation Learning via Dual Cost-Sensitive Graph Convolutional Network |
| title_short | Long-Tailed Graph Representation Learning via Dual Cost-Sensitive Graph Convolutional Network |
| title_sort | long tailed graph representation learning via dual cost sensitive graph convolutional network |
| topic | graph convolutional network imbalanced data classification cost-sensitive learning semi-supervised learning |
| url | https://www.mdpi.com/2072-4292/14/14/3295 |
| work_keys_str_mv | AT yijunduan longtailedgraphrepresentationlearningviadualcostsensitivegraphconvolutionalnetwork AT xinliu longtailedgraphrepresentationlearningviadualcostsensitivegraphconvolutionalnetwork AT adamjatowt longtailedgraphrepresentationlearningviadualcostsensitivegraphconvolutionalnetwork AT haitaoyu longtailedgraphrepresentationlearningviadualcostsensitivegraphconvolutionalnetwork AT stevenlynden longtailedgraphrepresentationlearningviadualcostsensitivegraphconvolutionalnetwork AT kyoungsookkim longtailedgraphrepresentationlearningviadualcostsensitivegraphconvolutionalnetwork AT akiyoshimatono longtailedgraphrepresentationlearningviadualcostsensitivegraphconvolutionalnetwork |