An attempt in modelling early intervention in autism using neural networks

We present a solution to a problem of early intervention in autistic learning. This is an addition to our model of autism which is based on Kohonen self-organizing maps extended with the source familiarity filter and the attention shift mechanism. In particular we study the feature map formation when attention shift is restricted by familiarity preference. The network learns the stimuli from the source with the lowest variability in great detail at the expense of the other source. The early intervention neural controller modifies the probabilities of presenting stimuli from a given source in response to the attention shift acceptance/rejection signals.