A New Model to Simulate the Formation of Orientation Columns Map in Visual Cortex

Abstract

The new model presented in the paper is used to simulate the development process of the orientation selectivity in primary visual cortex. The model combines mechanisms such as receptive field control, lateral connections, function columns into a network and then trained with random samples, can be regarded as a first stone of other feature maps. The model attempts to verify the basic features of the orientation map such as singularities, continuity and diversity. Meanwhile the model can be expanded with increasing columns or hyper-columns easily in order to process lager scope of stimulus. Another point is fault-tolerance, if some column is not successfully trained, the map can still perform well. After fast training process, the image of finished orientation map displaying with topology function is similar with the biological cortex orientation map, and the formed map can be used to extra the orientation information of the input quickly for the further visual process.