Unified Graph Networks (UGN): A Deep Neural Framework for Solving Graph Problems
Rudrajit Dawn, Madhusudan Ghosh, Partha Basuchowdhuri, Sudip Kumar Naskar · 2025
Read Unified Graph Networks (UGN): A Deep Neural Framework for Solving Graph Problems by Rudrajit Dawn, Madhusudan Ghosh, Partha Basuchowdhuri, Sudip Kumar Naskar (2025) 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-2502-07500.txt.
Description
Rudrajit Dawn and co-authors propose one neural framework for many graph problems instead of a bespoke model per task. Their 2025 paper introduces Unified Graph Networks, which blend graph convolution with ordinary image-style convolution. The goal mirrors what transformers did for text: a single adaptable engine across a whole problem family.
The unified model tackles link prediction, node classification, community detection, graph translation, and knowledge graph completion, among others. Small task-specific touches, like supernodes that shrink huge networks, preserve the shared core. Across twelve datasets and six unrelated problems, the framework beat specialist baselines on ten and matched on another. The authors argue graph learning has lacked its general-purpose moment. Their results suggest a unified design can compete with, and often beat, hand-built specialists.
Catalog record
| CREATOR | Rudrajit Dawn, Madhusudan Ghosh, Partha Basuchowdhuri, Sudip Kumar Naskar |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 55 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2502-07500