A Model of the Visual Dorsal Pathway for Computing Coordinate Transformations: An Unsupervised Approach

In humans, the problem of coordinate transformations is far from being completely understood. The problem is often addressed using a mix of supervised and unsupervised learning techniques. In this paper, we propose a novel learning framework which requires only unsupervised learning. We design a neural architecture that models the visual dorsal pathway and learns coordinate transformations in a computer simulation comprising an eye, a head and an arm (each entailing one degree of freedom). The learning is carried out in two stages. First, we train a posterior parietal cortex (PPC) model to learn different frames of reference transformations. And second, we train a head-centered neural layer to compute the position of an arm with respect to the head. Our results show the self-organization of the receptive fields (gain fields) in the PPC model and the self-tuning of the response of the head-centered population of neurons.