WEKO3
アイテム
Contextualized Messages Boost Graph Representations
http://hdl.handle.net/10061/0002001192
http://hdl.handle.net/10061/000200119247881294-2272-42d9-819f-b8297f8f6f8e
| アイテムタイプ | 学術雑誌論文 / Journal Article(1) | |||||||||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| 公開日 | 2025-10-06 | |||||||||||||||||
| タイトル | ||||||||||||||||||
| タイトル | Contextualized Messages Boost Graph Representations | |||||||||||||||||
| 言語 | ||||||||||||||||||
| 言語 | eng | |||||||||||||||||
| 資源タイプ | ||||||||||||||||||
| 資源タイプ | journal article | |||||||||||||||||
| アクセス権 | ||||||||||||||||||
| アクセス権 | open access | |||||||||||||||||
| 著者 |
Lim, Brian Godwin
× Lim, Brian Godwin
× Lim, Galvin Brice Sy
× Tan, Renzo Roel Perez
× 池田, 和司
|
|||||||||||||||||
| 抄録 | ||||||||||||||||||
| 内容記述タイプ | Abstract | |||||||||||||||||
| 内容記述 | Graph neural networks (GNNs) have gained significant attention in recent years for their ability to process data that may be represented as graphs. This has prompted several studies to explore their representational capability based on the graph isomorphism task. Notably, these works inherently assume a countable node feature representation, potentially limiting their applicability. Interestingly, only a few study GNNs with uncountable node feature representation. In the paper, a new perspective on the representational capability of GNNs is investigated across all levels—node-level, neighborhood-level, and graph-level— when the space of node feature representation is uncountable. Specifically, the injective and metric requirements of previous works are softly relaxed by employing a pseudometric distance on the space of input to create a soft-injective function such that distinct inputs may produce similar outputs if and only if the pseudometric deems the inputs to be sufficiently similar on some representation. As a consequence, a simple and computationally efficient soft-isomorphic relational graph convolution network (SIR-GCN) that emphasizes the contextualized transformation of neighborhood feature representations via anisotropic and dynamic message functions is proposed. Furthermore, a mathematical discussion on the relationship between SIR-GCN and key GNNs in literature is laid out to put the contribution into context, establishing SIR-GCN as a generalization of classical GNN methodologies. To close, experiments on synthetic and benchmark datasets demonstrate the relative superiority of SIR-GCN, outperforming comparable models in node and graph property prediction tasks. | |||||||||||||||||
| 書誌情報 |
en : Transactions on Machine Learning Research ページ数 19, 発行日 2025-04-07 |
|||||||||||||||||
| ISSN | ||||||||||||||||||
| 収録物識別子タイプ | EISSN | |||||||||||||||||
| 収録物識別子 | 2835-8856 | |||||||||||||||||
| 出版者版URI | ||||||||||||||||||
| 関連タイプ | isReplacedBy | |||||||||||||||||
| 識別子タイプ | URI | |||||||||||||||||
| 関連識別子 | https://openreview.net/pdf?id=sXr1fRjs1N | |||||||||||||||||
| 権利 | ||||||||||||||||||
| 権利情報Resource | https://creativecommons.org/licenses/by/4.0/ | |||||||||||||||||
| 権利情報 | All TMLR submissions, from the time of submission to final publication, are licensed under CC BY 4.0. At all times, copyright is retained by the authors. Authors are also allowed to upload their submissions to arXiv or other preprint servers at any time, either anonymously or including their identity, though double blind of the TMLR submission itself must be maintained by not linking to another version that includes the authors' names. | |||||||||||||||||
| 著者版フラグ | ||||||||||||||||||
| 出版タイプ | NA | |||||||||||||||||
| 助成情報 | ||||||||||||||||||
| 助成機関名 | Japan Society for the Promotion of Science (JSPS) | |||||||||||||||||
| 研究課題番号 | 18K19821 | |||||||||||||||||
| 研究課題番号URI | https://kaken.nii.ac.jp/grant/KAKENHI-PROJECT-18K19821/ | |||||||||||||||||
| 研究課題名 | 深層学習の理論解析による次世代脳型人工知能技術の開発 | |||||||||||||||||
| 助成情報 | ||||||||||||||||||
| 助成機関名 | Kyoto University and Toyota Motor Corporation |
|||||||||||||||||
| 研究課題名 | Advanced Mathematical Science for Mobility Society | |||||||||||||||||
