A stochastic model of retinotopy: A self organizing process (Q1084351)

In this paper the retina R is modelled as a two-dimensional network \(\{0,1,...,n+1\}^ 2\) (and also as an integer interval \(\{0,1,...,N+1\})\). Each cell (i,j) is mainly related to a cortex cell \(x_{ij}\) to which a density of synaptic weights \(f_{ij}(x)\) is associated. It is assumed that \(f_{ij}(x)=f(\| x-x_{ij}\|)\) where f is decreasing, positive and \(f(0)=1\). It is assumed that the system is subject to stimulation chosen to render the resulting model Markovian in the one-dimensional case. The convergence of the corresponding Markov chain to a stationary distribution is established which in turn establishes the retinotopy of the system. The difficulty in extending the results to the two- dimensional case is pointed out; in this case the asymptotic behaviour is brought out by Monte-Carlo simulation.