Emergent symmetry of local and global maps in the primary visual cortex: Self-organisation of orientation preference.

Abstract

We present a model of the cellular organization in the primary visual cortex which is based upon the idea that there are two visuotopic mappings, one global and the other local. The local visuotopic maps self-organize from a set of initially random inputs. The various response properties measured in V1, and the regular geometrical relationships between them, are explained in terms of the interaction of these two mappings. We describe computer modelling of orientation preference in V1 which relies on two assumptions: 1) the receptive fields develop due to a simple excitatory-centre / inhibitory-surround mechanism, and 2) any point in the global visuotopic map can reach any point in the non-granular layers via poly-synaptic routes. Hebbian learning is applied to these indirect inputs while the network is stimulated with a moving bar. The computer modelling shows that all the key geometrical features of orientation preference; singularities, linear zones and saddle-points, emerge consequent to the learning. This suggests that orientation preference is a by-product of the double visuotopic mapping. More generally, the two mappings are hypothesized to allow those features of the visual field which tend to be spatially contiguous in the visual field (orientation, texture, colour, contrast) to be learned as response properties by neurons within V1.