You are here: Home Talks & Events Bernstein Seminar 2011 Sen Cheng (Mercator Research Group "Structure of Memory", Ruhr-University Bochum, Germany)

Sen Cheng (Mercator Research Group "Structure of Memory", Ruhr-University Bochum, Germany)

— filed under:

"From grid cells to place cells: A generic and robust principle" / Tuesday, February 15, 2011, 17:15 h

  • Bernstein Seminar
When Feb 15, 2011
from 05:15 PM to 07:20 PM
Where Lecture Hall, Hansastr. 9a
Add event to calendar vCal









The Bernstein Center Freiburg

Bernstein Seminar
Sen Cheng
Mercator Research Group "Structure of Memory"
Ruhr-University Bochum, Germany

From grid cells to place cells:
A generic and robust principle

Tuesday, February 15, 2011

17:15 h

Lecture Hall (ground floor)
Bernstein Center Freiburg
Hansastraße 9A
79104 Freiburg
How do spatial representations in the hippocampus arise? The discovery of grid cells in the medial entorhinal cortex split this question into two:

1. How are periodic grids generated?
2. How is this periodic input transformed into the punctate place fields downstream?

In this talk, I will focus on the second question. Several models were proposed to explain the grids-to-places transformation, however, these prior studies have three limitations. First, the models do not provide a solution for the grids-to-places transformation given a set of grids and multiple place fields in arbitrary locations. Second, the proposed models support only one spatial map whereas experimental data suggest that place cells might support multiple maps. Third, it remains unclear what exactly, if anything, the various previous models have in common. I will address these limitations. I will derive a simple solution for the synaptic weights in a feedforward network to generate the grids-to-places transformation. This relationship is highly robust to noise in the inputs and synaptic weights. As a result, weights that look noisy in one context contribute to strong spatial specificity in another context, and vice versa. I will use this insight to generate multiple spatial maps with place fields in any arbitrarily-given location. Finally, the same principle emerges in the solutions of all models studied, except for one that is highly unrealistic.
The talk is open to the public. Guests are cordially invited!
Personal tools