Cooperative organization of connectivity patterns and receptive fields in the visual pathway: Application to adaptive tresholding

Abstract

A biologically plausible theoretical framework to embody cooperative computation in the visual pathway is proposed. From photoreceptors to ganglion cells, visual processing is” properly interpreted by means of linear and nonlinear spatio-temporal filters with center-periphery receptive fields and analogic computation. At cortical level (mainly recurrent pyramidals) hybrid formulations using a combination of local operators (sum plus sigmoid) mid conditionals” (inferential rules) are more appropriate. This” inferential model is quite general, supports analogic and logic computation as particular cases, and should be applicable to bridge the gap between connectionistic and symbolic artificial intelligence in general and between low level and high level vision, in particular. To illustrate the possibilities of the model, tapographic reorganization of connectivity patterns and receptive fields are considered. Adaptive thresholding as a consequence of cooperative consensus on homogeneity measures in the neighbourhood of each neuron has been simulated Other properties such as self-organization of columns of contrast, orientation, speed or preferred direction can also be modelled as” cooperative processes.