GraphLCP : Structure-Aware Localized Conformal Prediction on Graphs
Peyman Baghershahi, Fangxin Wang, Debmalya Mandal, Sourav Medya · 2026
Read GraphLCP : Structure-Aware Localized Conformal Prediction on Graphs by Peyman Baghershahi, Fangxin Wang, Debmalya Mandal, Sourav Medya (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-2605-08074.txt.
Description
Peyman Baghershahi, Fangxin Wang, Debmalya Mandal, and Sourav Medya bring conformal prediction to graph neural networks. Conformal methods give prediction sets with finite-sample guarantees, but graphs make localization tricky. Judging closeness by embedding distance alone can be unreliable and can produce sets that are too large to help.
GRAPHLCP builds location and weighting from the graph structure itself. A feature-aware densification step reduces bias in sparse regions, and a Personalized PageRank kernel measures structural closeness between nodes. Anchor sampling and calibration weights then follow both local and long-range links, aiming for sharper prediction sets that still keep their coverage guarantee on graph data.
Catalog record
| CREATOR | Peyman Baghershahi, Fangxin Wang, Debmalya Mandal, Sourav Medya |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 64 KB |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2605.08074