Jonathan Victor (Department of Neurology and Neuroscience, Weill Cornell Medical College, New York, USA)

The Bernstein Center Freiburg

Bernstein Seminar

Jonathan D. Victor

Department of Neurology and Neuroscience
Weill Cornell Medical College, New York, USA

 

Understanding the Computations
in Primary Visual Cortex:
Does Tweaking the Standard Model Suffice?

Abstract:
A central problem in systems neuroscience is to understand the nature of computations by neural networks. The visual system is an excellent model for addressing this question, and our knowledge of the retina illustrates the kind of success we can hope to achieve. Our understanding of cortical computations is at present much more limited. The standard model consists of a bank of feedforward filters and simple nonlinearities. However, there are substantial differences between the computations performed by real cortical neurons and computations of models based on a feedforward cascade.
These differences are present not only for responses to complex real-world natural images, but, as we show, when visual cortex is probed with artificial stimuli with a relatively simple mathematical structure. The common denominator in these stimuli is the presence of high-order spatial correlations. These observations suggest that a strongly recurrent network is an appropriate basic framework for understanding cortical computations.

Host: Michael Bach