CORTICAL ORIENTATION MAP DEVELOPMENT FROM NATURAL IMAGES: THE ROLE OF CORTICAL RESPONSE AMPLIFICATION IN V1 CHRISTIAN PIEPENBROCK and KLAUS OBERMAYER

Simple cells in the primary visual cortex respond selectively to oriented stimuli. It has been proposed that such feature detecting neurons should generate a sparse representation of the visual world and orientation selective receptive fields are in this sense optimal spatial filters for “natural” visual environments. In this contribution we show that a competitive Hebbian development model driven by natural images may generate a topographic projection from the lateral geniculate nucleus to the primary visual cortex as well as orientation selective receptive fields. The resulting representation is sparse and the degree of sparseness depends on the recurrent dynamics, i.e., the lateral competition among the cortical neurons. Simulations show that for weak competition, the resulting receptive fields are global and unstructured and for intermediate competition, they refine and a topographic projection emerges. Finally, in the case of strong competition, the receptive fields act as oriented localized spatial filters arranged in a typical orientation selectivity map.