Self-organization of the velocity selectivity of directionally selective cells

A self-learning algorithm is proposed that can develop the velocity selectivity of directionally selective cells. This learning algorithm is simple in that it can be described only with the presynaptic and postsynaptic potentials. We introduce the algorithm for a model called a " mass model" that is constructed by using the basic network which can detect specific direction and velocity. Numerical simulation results show that each of the basic network in the mass model learns to have the selectivity for different optimum velocity.