Abstract
A model of the self-organization of synapses in the visual cortex is presented. Subject to Hebbian learning with decay, evolution of synaptic strengths proceeds to a stable state in which all synapses have either maximum, or minimum, pre/post-synaptic coincidence. The most stable configuration gives rise to anatomically realistic “local maps”, each of macro-columnar size, and each organized as Mobius projections of retinotopic space. A tiling of V1, constructed of approximately mirror-image reflections of each local map by its neighbors is formed, accounting for orientation-preference singularities, linear zones, and saddle points - with each map linked by connections between sites of common orientation preference. Ocular dominance columns, the occurrence of direction preference fractures always in odd numbers around singularities, and effects of stimulus orientation relative to velocity of motion, are accounted for. Convergence to this configuration is facilitated by the spatio-temporal learning rule.