The limbic navigation system as a hierarchical RNN
- 1UCL
Abstract
Path-integration-trained recurrent neural networks (RNNs) can reproduce grid-like and head-direction-like representations, but most existing models rely on low-dimensional velocity inputs and do not capture the hierarchical organisation of the brain's self-localisation circuit. We introduce a hierarchical recurrent neural network (HRNN) that separately processes angular velocity and linear speed to perform simultaneous angular and positional path integration. Trained on these tasks, the HRNN generalises to longer timescales than seen during training and develops neural response properties resembling all the key cell types: angular velocity cells, head direction cells, speed cells, conjunctive and pure grid cells. The learned circuit also recovers key motifs of the navigation system, including attractor dynamics and asymmetric connectivity patterns. Perturbation analyses reveal functional information flow across the hierarchy, consistent with experimental observations. Finally, introducing theta-paced mutual inhibition between two learned angular-velocity subpopulations enables the HRNN to reproduce the bidirectional theta sweeps observed in empirical data. These results establish hierarchical task-driven RNNs as a biologically interpretable framework for linking neural representations, circuit motifs, and population dynamics in spatial navigation, while enabling targeted in silico perturbations and the generation of experimentally testable predictions.
Related
- Inverse covariation between the neocortex and the limbic system is concentrated at deep evolutionary divergences
- A hierarchical navigation decision-making method for UAV swarms in unknown communication-constrained environments
- An inter-row navigation system design for a safflower picking robot
- A Hierarchical OhrR-EmrR Regulatory Circuit Fine-Tunes the Vfm Quorum- Sensing System and Virulence in Dickeya oryzae.
- The development of navigation and attentional abilities shapes how children explore naturalistic environments
BibTeX
@misc{ji2026limbic,
title = {{The limbic navigation system as a hierarchical RNN}},
author = {Z. Ji and H. Zhang and K. Gorantla and N. Burgess},
howpublished = {bioRxiv},
year = {2026},
month = oct,
doi = {10.64898/2026.10.19.760224},
url = {https://www.biorxiv.org/content/10.64898/2026.10.19.760224v1},
note = {CC BY},
}
RIS
TY - UNPB
TI - The limbic navigation system as a hierarchical RNN
AU - Z. Ji
AU - H. Zhang
AU - K. Gorantla
AU - N. Burgess
PY - 2026
DA - 2026/10/19
PB - bioRxiv
DO - 10.64898/2026.10.19.760224
UR - https://www.biorxiv.org/content/10.64898/2026.10.19.760224v1
AB - Path-integration-trained recurrent neural networks (RNNs) can reproduce grid-like and head-direction-like representations, but most existing models rely on low-dimensional velocity inputs and do not capture the hierarchical organisation of the brain's self-localisation circuit. We introduce a hierarchical recurrent neural network (HRNN) that separately processes angular velocity and linear speed to perform simultaneous angular and positional path integration. Trained on these tasks, the HRNN generalises to longer timescales than seen during training and develops neural response properties resembling all the key cell types: angular velocity cells, head direction cells, speed cells, conjunctive and pure grid cells. The learned circuit also recovers key motifs of the navigation system, including attractor dynamics and asymmetric connectivity patterns. Perturbation analyses reveal functional information flow across the hierarchy, consistent with experimental observations. Finally, introducing theta-paced mutual inhibition between two learned angular-velocity subpopulations enables the HRNN to reproduce the bidirectional theta sweeps observed in empirical data. These results establish hierarchical task-driven RNNs as a biologically interpretable framework for linking neural representations, circuit motifs, and population dynamics in spatial navigation, while enabling targeted in silico perturbations and the generation of experimentally testable predictions.
ER -