Modeling of orientation preference map construction

Emre Baspinar (Inria, MathNeuro)

Mardi 7 mai, 11h00, salle Coriolis (Galois).

Abstract. In this talk, I will first present a geometric framework describing the functional architecture of the primary visual cortex (V1). This framework is based on an anisotropic, degenerate geometry known as sub-Riemannian geometry. Then, we will see a model which was extended from this sub-Riemannian geometry, and which can generate the orientation preference maps of the neurons found in V1. The orientation maps are 2D representations of the orientation angle values which the V1 neurons are sensitive to. These maps are known to be constructed by the brain up to a certain level before the eye opening in the early natal period. Our model considers a noise image as the input stimulus inducing the orientation map construction. It follows a biologically inspired method to generate the orientation map by starting from this noise image.