The self-organization of models of direction selective cells in visual cortex

Direction-selective cells in primary visual cortex in vivo show prominent response only to the visual stimuli moving toward a certain direction. Physiological data about such direction-selective cells suggest that the action of direction selective cell depend on proper cooperation of excitatory and inhibitory inputs. When only inhibitory inputs are blocked pharmacologically, characteristic phase transition (iceberg effect) emerges. In this paper, I referred to inhibitory system responsible for direction selectivity from physiological data, and I propose neural network models of direction selective cells. In these models the distributed excitatory system and feedforward lateral inhibitory system cooperate with each other. One of these models shows the same phase transition as real direction-selective cells show when only inhibitory system are blocked. Next, I propose the method to self-organize direction selectivity in these models by controlling both the lateral inhibitory system and excitatory system separately, obeying different learning rules. By using the method, stable self-organization of direction selective cell models can be developed.