Recurrent Graph Neural Networks and Arithmetic Circuits

Timon Barlag, Vivian Holzapfel, Laura Strieker, Jonni Virtema, Heribert Vollmer · 2026

Read Recurrent Graph Neural Networks and Arithmetic Circuits by Timon Barlag, Vivian Holzapfel, Laura Strieker, Jonni Virtema, Heribert Vollmer (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-2603-05140.txt.

Description

Timon Barlag and colleagues prove an exact match between recurrent graph neural networks and recurrent arithmetic circuits over the real numbers. Their circuits use memory gates to store data between iterations, taking labelled graphs encoded as tuples of real numbers and computing the same function as the networks. In the other direction, recurrent networks fed circuit inputs as starting feature vectors reproduce circuit outputs among their node features. The correspondence opens a circuit-complexity lens on trained recurrent networks.

Catalog record

CREATORTimon Barlag, Vivian Holzapfel, Laura Strieker, Jonni Virtema, Heribert Vollmer
YEAR2026
COLLECTIONPAPERS
FORMATTXT
FILE SIZE101788
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2603.05140

Browse the full library · About the Tower of Babel