Author : Harry G. Barrow
Publisher :
ISBN 13 :
Total Pages : 32 pages
Book Rating : 4.:/5 (352 download)
Book Synopsis Models of Synaptic Development in Early Visual Cortex by : Harry G. Barrow
Download or read book Models of Synaptic Development in Early Visual Cortex written by Harry G. Barrow and published by . This book was released on 1996 with total page 32 pages. Available in PDF, EPUB and Kindle. Book excerpt: Abstract: "This report is comprised of three independent parts describing developmental models for orientation selectivity, 'colour-blobs' and position-invariant complex cells respectively. Each of these models has been documented, in abbreviated form, in previously published work [1,2,3]. As such, these reports are intended to present 'the full story'. Part I describes an activity-based model of processing in the early visual pathway including retina, lateral geniculate nucleus (LGN), and simple cells in the cortex. We use a non-adaptive model of on- and off-centre retinal ganglion cells which project to the LGN. Cortical simple cells receive excitatory projections from the LGN. There are two populations of cortical cells: one population excites close neighbours whilst the other inhibits all neighbours within a greater area. When presented with input the network activity settles, after which feedforward weights connecting the simple and geniculate cells adapt using a Hebbian rule. After many presentations from natural images the feedforward weights adapt to become 'edge' and 'bar' detectors, with many receptive fields being similar to Gabor functions. These are orientation selective, and orientation preference varies smoothly across the cortical surface. Part II describes a model of colour-processing in the parvicellular visual pathway of primates. Principal component analysis of many small spatial samples taken from a natural colour image shows that only one of the eigenvectors with a large eigenvalue is significantly colour selective, and this eigenvector is not selective for 2D orientation. This result suggests an activity- based explanation for the formation of 'colour-blobs' in layer IVb of primate striate cortex. The work describes a network simulation of processing in the primate retina, LGN and V1 which self-organises in response to natural colour images and produces 'feature-maps' in which islands of a few colour-sensitive cells (i.e. colour blobs) are surrounded by a sea of oriented non-colour-selective cells. Part III proposes that complex cells in layers II & III of primate visual cortex learn to be sensitive to orientation, but invariant to position, through a mechanism similar to classical conditioning. A simulation is described with a simple computational model which demonstrates how complex cells can develop strong connections to simple cells with similar orientation preferences at different spatial positions by using a temporal version of the Hebbian learning rule. A further simulation with a detailed biological model of the primate visual pathway from retina to cortex supports the claim that complex cells learn orientation-selectivity but spatial invariance using temporal Hebbian learning."