Statistical physics analysis of graph neural networks: Approaching optimality in the contextual stochastic block model
Duranthon, Lenka Zdeborová · 2025
Read Statistical physics analysis of graph neural networks: Approaching optimality in the contextual stochastic block model by Duranthon, Lenka Zdeborová (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-2503-01361.txt.
Introduction
Duranthon and Lenka Zdeborová wrote this 2025 theory paper on graph neural networks. These tools learn from data arranged as networks, but why they work well is still poorly understood. One known flaw is oversmoothing, where data from far-apart nodes gets blurred together after many steps.
The pair study a simple form of the network trained to sort nodes into groups. Using tools from statistical physics, they work out how well the network should do as the data grows large. Their math shows that going deeper can bring the network close to the best possible score. They also show how the design must scale with depth to avoid the blurring flaw.
Catalog record
| CREATOR | Duranthon, Lenka Zdeborová |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 160 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2503-01361