Geometric constraints and cognitive inputs jointly shape emergent brain dynamics and topology
- 1Department of Psychiatry, Brain Health Institute, Rutgers University, Piscataway, NJ, USA
Abstract
How the brain's physical geometry gives rise to its flexible functional repertoire remains a central question in neuroscience. Here, we approached this question through recurrent neural networks (RNNs). We trained three RNN classes on a cognitive working-memory task, forming a graded hierarchy of spatial constraints: Vanilla RNNs (no spatial constraints), Masked RNNs (projection constraints limiting where information enters and leaves the network), and biophysical RNNs (bioRNNs; projection constraints and spatial embedding of the networks' connectivity using the brain's inter-regional Euclidean geometry). We assessed how well each RNN class predicted empirical fMRI activity without exposing them to it during training. Our results showed that bioRNNs were the only networks to successfully predict empirical brain activity and to organize their dynamics into a spatial pattern that recapitulated the brain's principal hierarchy (the sensorimotor-association axis). Additionally, bioRNNs' ability to predict empirical brain activity emerged along a training trajectory in which the effects of brain geometry were encoded by the network first, then partly traded back as the task was mastered. Importantly, brain-like topological features emerged in bioRNNs as they increased their task proficiency while maintaining their ability to predict brain activity. Taken together, our results indicate that physical geometry and cognitive inputs play distinct, complementary roles: while geometry constrains the space of possible brain dynamics, cognitive inputs determine which dynamics are expressed. Our results also situate topology as the scaffold through which the physically embedded brain reconciles wiring costs and computational demands.
Related
- Stochasticity in Metabolism: Oxidative Stress Response Phenotypes As Emergent Consequences of Network Topology
- Bridging Sensory Extraction and Decision Dynamics: Face Pareidolia Emergent from Bottom-Up and Top-Down Interactions
- Higher-order topology uncovers the multidimensional dynamics of naturalistic emotion
- Giving particles legs: How balance recovery may drive emergent collective dynamics in ultra-dense crowds
- Multiscale Topology-Regulated Nonlinear Dynamics for Protein B-Factor Prediction and Single-Cell Type Classification
BibTeX
@misc{beyh2026geometric,
title = {{Geometric constraints and cognitive inputs jointly shape emergent brain dynamics and topology}},
author = {A. Beyh and J. Z. Kim and W. U. Bajwa and L. Parkes},
howpublished = {bioRxiv},
year = {2026},
month = aug,
doi = {10.64898/2026.08.07.742341},
url = {https://www.biorxiv.org/content/10.64898/2026.08.07.742341v3},
note = {CC BY},
}
RIS
TY - UNPB
TI - Geometric constraints and cognitive inputs jointly shape emergent brain dynamics and topology
AU - A. Beyh
AU - J. Z. Kim
AU - W. U. Bajwa
AU - L. Parkes
PY - 2026
DA - 2026/08/07
PB - bioRxiv
DO - 10.64898/2026.08.07.742341
UR - https://www.biorxiv.org/content/10.64898/2026.08.07.742341v3
AB - How the brain's physical geometry gives rise to its flexible functional repertoire remains a central question in neuroscience. Here, we approached this question through recurrent neural networks (RNNs). We trained three RNN classes on a cognitive working-memory task, forming a graded hierarchy of spatial constraints: Vanilla RNNs (no spatial constraints), Masked RNNs (projection constraints limiting where information enters and leaves the network), and biophysical RNNs (bioRNNs; projection constraints and spatial embedding of the networks' connectivity using the brain's inter-regional Euclidean geometry). We assessed how well each RNN class predicted empirical fMRI activity without exposing them to it during training. Our results showed that bioRNNs were the only networks to successfully predict empirical brain activity and to organize their dynamics into a spatial pattern that recapitulated the brain's principal hierarchy (the sensorimotor-association axis). Additionally, bioRNNs' ability to predict empirical brain activity emerged along a training trajectory in which the effects of brain geometry were encoded by the network first, then partly traded back as the task was mastered. Importantly, brain-like topological features emerged in bioRNNs as they increased their task proficiency while maintaining their ability to predict brain activity. Taken together, our results indicate that physical geometry and cognitive inputs play distinct, complementary roles: while geometry constrains the space of possible brain dynamics, cognitive inputs determine which dynamics are expressed. Our results also situate topology as the scaffold through which the physically embedded brain reconciles wiring costs and computational demands.
ER -