Saved in:
Bibliographic Details
Main Authors: Allier, Cédric, Heinrich, Larissa, Schneider, Magdalena, Saalfeld, Stephan
Format: Preprint
Published: 2026
Subjects:
Online Access:https://arxiv.org/abs/2602.13325
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866915797902819328
author Allier, Cédric
Heinrich, Larissa
Schneider, Magdalena
Saalfeld, Stephan
author_facet Allier, Cédric
Heinrich, Larissa
Schneider, Magdalena
Saalfeld, Stephan
contents Graph neural networks trained to predict observable dynamics can be used to decompose the temporal activity of complex heterogeneous systems into simple, interpretable representations. Here we apply this framework to simulated neural assemblies with thousands of neurons and demonstrate that it can jointly reveal the connectivity matrix, the neuron types, the signaling functions, and in some cases hidden external stimuli. In contrast to existing machine learning approaches such as recurrent neural networks and transformers, which emphasize predictive accuracy but offer limited interpretability, our method provides both reliable forecasts of neural activity and interpretable decomposition of the mechanisms governing large neural assemblies.
format Preprint
id arxiv_https___arxiv_org_abs_2602_13325
institution arXiv
publishDate 2026
record_format arxiv
spellingShingle Graph neural networks uncover structure and functions underlying the activity of simulated neural assemblies
Allier, Cédric
Heinrich, Larissa
Schneider, Magdalena
Saalfeld, Stephan
Neurons and Cognition
Machine Learning
Graph neural networks trained to predict observable dynamics can be used to decompose the temporal activity of complex heterogeneous systems into simple, interpretable representations. Here we apply this framework to simulated neural assemblies with thousands of neurons and demonstrate that it can jointly reveal the connectivity matrix, the neuron types, the signaling functions, and in some cases hidden external stimuli. In contrast to existing machine learning approaches such as recurrent neural networks and transformers, which emphasize predictive accuracy but offer limited interpretability, our method provides both reliable forecasts of neural activity and interpretable decomposition of the mechanisms governing large neural assemblies.
title Graph neural networks uncover structure and functions underlying the activity of simulated neural assemblies
topic Neurons and Cognition
Machine Learning
url https://arxiv.org/abs/2602.13325