Graph Domain Adaptation via Homophily-Agnostic Reconstructing Structure
Ruiyi Fang, Shuo Wang, Ruizhi Pu, Qiuhao Zeng, Hao Zheng, Ziyan Wang, Jiale Cai, Zhimin Mei, Song Tang, Charles Ling, Boyu Wang · 2026
Read Graph Domain Adaptation via Homophily-Agnostic Reconstructing Structure by Ruiyi Fang, Shuo Wang, Ruizhi Pu, Qiuhao Zeng, Hao Zheng, Ziyan Wang, Jiale Cai, Zhimin Mei, Song Tang, Charles Ling, Boyu Wang (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2602-07573.txt.
Description
Graph domain adaptation moves knowledge from labeled source graphs to unlabeled target graphs, but most methods assume both graphs are homophilic, where neighbors share labels. Ruiyi Fang and ten co-authors drop that assumption with a homophily-agnostic method that handles varying homophily on both sides.
A divide-and-conquer strategy reconstructs highly homophilic and heterophilic variants of source and target graphs separately, then aligns knowledge between matching variants. Experiments on five benchmarks show superior results where homophily varies. Graph learning teams get adaptation that works whether neighbors agree or disagree, a common reality the field long assumed away.
Catalog record
| CREATOR | Ruiyi Fang, Shuo Wang, Ruizhi Pu, Qiuhao Zeng, Hao Zheng, Ziyan Wang, Jiale Cai, Zhimin Mei, Song Tang, Charles Ling, Boyu Wang |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 49 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2602-07573