Theoretical and Computational Neuroscience
Author: Federico Szmidt Szmidt | email: fszmidt@gmail.com
Federico Szmidt 1°, Camilo J. Mininni 1°
1° Instituto de Biología y Medicina Experimental – CONICET
The entorhinal cortex is a key brain region regarding the neural representation of space and time. Neurons in the medial (MEC) and lateral (LEC) entorhinal cortex respectively show a preference for coding the animals position or the passage of time, although some mixed selectivity can also be found. Several models have been proposed that explain the emergence and mechanisms behind space and time coding. However, how the representations interact within these interconnected regions is not well understood. In this work we trained recurrent neural networks to encode an agents position, by integrating speed from simulated trajectories, together with the elapsed time between discrete stimuli presentations. We studied the emerging codes, comparing them between networks, and with results from experiments in behaving animals. Trained networks exhibited neurons that coded position by firing in a quasi periodic fashion, reminiscent of grid cells in MEC, and neurons with firing rates that correlated with elapsed time since the last stimulus, reminiscent of ramping activity as reported in LEC. Most neurons specialized in coding either space or time, conforming two separated populations, with a reduced number of neurons with mixed selectivity. Our results suggest that independent codes for space and time are the expected solution to a joint space-time optimization problem.