Topology-Aware Gaussian Graph Repair for Robust Graph Neural Networks Journal: Expert Systems with Applications
Anubha Goel, Juho Kanniainen · 2026
Read Topology-Aware Gaussian Graph Repair for Robust Graph Neural Networks Journal: Expert Systems with Applications by Anubha Goel, Juho Kanniainen (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-2606-03462.txt.
Description
Goel and Kanniainen repair flawed graphs before graph neural networks learn from them. Real graphs are often imperfect in two ways. Noisy edges can link nodes that do not belong together. Missing edges can block useful signals from spreading. Past fixes mostly delete suspect edges or rebuild the graph during training. Deletion cannot restore lost links, and rebuilding adds heavy cost.
Their method, TAGR, builds a sparse graph of feature neighbors with an adaptive Gaussian kernel. Those added edges join nodes with similar features. A residual step then keeps and reweights the original edges by local consistency. The repaired graph feeds standard networks with no change to their design. Tests on citation benchmarks cover both noisy-edge and missing-edge damage. The Gaussian repair brings the main gain in robustness. The residual step adds stability when the graph is incomplete. Light sparse repair beats dense graph learning here.
Catalog record
| CREATOR | Anubha Goel, Juho Kanniainen |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 70 KB |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2606.03462