ePoster, COSYNE 2025
Coordinating control and planning for navigation on simplicial complex attractors
- Montreal, Canada
- Neuroscience, Computational Neuroscience and
Abstract
Animals learn complex cognitive maps to navigate through physical environments or reason about abstract concepts. These maps are believed to be encoded by mammalian hippocampal grid and place cells, forming low-dimensional neural manifolds that support flexible navigation. However, beyond simple line and ring attractors, how the brain represents and navigates more complex neural manifolds remains unclear. In this study, we propose a solution for representing arbitrary nonlinear neural manifolds by approximating them with piecewise linear spaces composed of simplexes, known as “simplicial complexes.” Our circuit represents these simplicial complexes as stable attractors and supports navigation through coordinated control and planning. First, we derive the synaptic weights and plasticity rules needed to embed the dynamics of a simplex attractor and enable closed-loop feedback control for navigation within it. Next, we mathematically identify the conditions required to stitch simplex attractors together, forming a cohesive simplicial complex attractor. Specifically, two simplex attractors can be connected through sufficient mutual inhibition. Furthermore, we design a circuit that navigates across simplexes using shortest-path planning via recurrent dynamics on a hypergraph that encodes the connectivity between simplexes. Finally, we map our circuit to specific components in the navigation circuits of rodents and Drosophila, offering a plausible mechanism for goal-directed navigation in complex environments and generating experimental predictions for both systems.
Topics
- cognitive maps, search World Wide
- place cells, search World Wide
- neural manifolds, search World Wide
- simplicial complexes, search World Wide
- simplex attractors, search World Wide
- synaptic weights, search World Wide
- plasticity rules, search World Wide
- shortest-path planning, search World Wide
- goal-directed navigation, search World Wide