Connectionists: Decoding grid cells

[Martin Stemmler]

Dear connectionists,

  We’d like to announce an electronic preprint for a paper on how the activity of grid cells in 
medial entorhinal cortex can be read out. This read-out relies on a  simple population vector average to yield ego-centric homing vectors.

http://dx.doi.org/10.1101/021204

Grid cells with lattices having multiple scales act in concert like the hour, minute, and second hands on an analog clock.
Except that for mammalian navigation,  the “clock” for space lies on a torus.

We show that the population vector average yields the maximum likelihood solution for dead reckoning
(exactly in 1D, and to a very good approximation in the non-trivial extension to 2D). Population-vector decoding also predicts
the discrete, modular arrangement of grid length scales, which form a geometric progression with a scale ratio of roughly 3/2.

best regards,

Martin Stemmler, Alexander Mathis, and Andreas Herz