• Home
• Staff & Students
• Vacancies

Elad Schneidman

 

Weizmann Institute of Science

 

(www.weizmann.ac.il/neurobiology/schneidman)

 

Wednesday 28th March 2012

4pm

 

B10 Seminar Room, Basement,

Alexandra House, 17 Queen Square, London, WC1N 3AR

 

 

Sparse high-order interaction networks underlie learnable neural population codes

 

 

Abstract:

Information is carried in the brain by the joint activity patterns of large groups of neurons. Understanding the structure and content of neural population codes is challenging, due to the exponential number of possible activity patterns and dependencies among neurons. Studying groups of 100 retinal neurons responding to natural movies, we found that they are strongly correlated and that pairwise maximum entropy models, which are highly accurate in describing small networks, are no longer sufficient. We show that because of the sparse nature of the neural activity, a very sparse high-order interaction network underlies the population code, which cab be learned easily with extremely high accuracy using a novel pseudo-likelihood model. We further show that the interaction network is organized in a hierarchical and modular manner, reflecting the scalability of the code, and that it can be used for decoding. Our results thus suggest that learnability is a key feature of the neural code.