Manifold learning for place cells, grid cells, and other neural systems

Summary

Date: 
March 30, 2017 - 11:00am
Location: 
Northwest Building 243
About the Speaker
Name: 
Sam Lewallen
Speaker Title: 
Postdoctoral Research Associate
Speaker Affiliation: 
Princeton Neuroscience Institute

With the advent of many new technologies for recording simultaneously from large
populations of neurons, there is an opportunity for new analysis methods to
help organize and interpret high-dimensional neural data. One appealing
approach is to think geometrically, and try to characterize the shape of the
data in neural activity space. In this talk, I'll describe our ongoing work to
develop and apply manifold learning techniques to study the geometry and
topology of neural activity, including techniques for directly incorporating
neural dynamics. Our analysis was initially inspired by the "neural
manifolds" of grid cells and place cells, and I hope to describe our
preliminary application to new calcium imaging data from the mouse medial
entorhinal cortex. This work is a joint collaboration with Ryan Low and Ben
Scott.