Computational properties and auto-organization of a population of cortical neurons

The validity of population coding for movement within the motor cortex has now been confirmed by many experimental studies. The preferred direction of the population units seemed to be uniformly distributed. Theoretical studies have shown that such population regularities are important to perform optimal learning with Hebb-like learning rules. This paper shows that population regularity can result from an auto-organization process driven by the difference between feedforward and lateral inputs, whatever the distribution of inputs in the case of a linear encoding model of the arm movement command