Theoretical and Computational Neuroscience
Author: Facundo Emina | Email: facuemina@gmail.com
Facundo Emina 1°2°, Sabrina Benas 1°2°, Ximena Fernandez 3°, Emilio Kropff 2°
1° Universidad de Buenos Aires. Facultad de Ciencias Exactas y Naturales. Departamento de Física, Buenos Aires, Argentina.
2° Fundación Instituto Leloir – IBBA/CONICET, Buenos Aires, Argentina.
3° Department of Mathematics, Durham University, UK.
Entorhinal grid cells use a hexagonal pattern to encode spatial information based on an animal’s location. Maps within a grid cell module share spacing and orientation, differing only in relative two-dimensional spatial phases. Feed-forward networks model grid pattern formation, with entorhinal cells assimilating spatial inputs through Hebbian plasticity. Alignment can be achieved if a two-dimensional continuous attractor recurrent connectivity is imposed. Yet, this architecture has the drawbacks of being complex to construct and rigid, allowing no deviations from the hexagonal pattern such as the ones experimentally observed.
Our study proposes a simpler approach: a one-dimensional attractor for grid alignment. Employing topological data analysis, population activity constitutes a torus-like sample, retaining essential architectural features. Contrary to convention, in our model, architecture and attractor representation aren’t topological identical entities, challenging prior assumptions of brain-wide attractor networks.
We also explore the possibility that one-dimensional attractors could enable path integration computations by harmonizing feed-forward and recurrent inputs along a 1D track. We discuss how extending these ideas to an open field exploration would suggest that entorhinal networks are not specifically tuned to perform 2D path integration but rather benefit from a simpler connectivity scheme which enables them to perform more flexible computations.