[The modelling of the process of image-sequence fixation and reproduction in a projective-associative network made up of excitatory neuron-like elements]

Abstract

On mathematic model of several interconnected networks of excitatory neurone-like elements realized in the form of program on computer "Nord-100", a study of conditions of fixation and reproduction of symbols (words) succession was conducted. Connections between the receptive (C1 and C2) and associative (A1 and A2) networks were by the principle "one to one", connections between the networks A1 and A2 with reinforcing general activating network (GAN) were by the principle "all with all". Possibility was shown of restoration of images succession fixed in the network on the basis of the principle of chain conditioned reflexes provided a successive change of reinforcing GAN elements by means of the decrease of the threshold of their activation. It was found that contacts transferring the influences of the reinforcing network at learning by the Hebb principle, must either initially exert a subthreshold action or be "unlearning" for the elimination of the process of overexcitation.