From a complex neural network with many components to an aggregated simpler neural network

We consider complex neural networks NN, with many similar NNC components with "interne" couplings of their neurones stronger than the "externe" couplings. Quasi-independent local activation in various NNC can be specified by the "interne" couplings. The "externe" couplings serving especially to the transfer of such local processes - after a corresponding delay - in a new NNC. According to previous simulations, we carl simplify the approximate analysis of such complex NN transforming them in simpler aggregated neural network ANN, using two aggregation rules: a) of the neurones of similar NNC in standardized "generalized-neurones", GN, with similar activations and b) of the "inter-component" couplings in "GN" standardized couplings. Thus, in ANN we can specify the excitation of the whole network, by adequate adaptations of: a) the reccurence relations - describing the overall activation propagation - and of b) the back-propagation algorithm - specifying the dynamical parameters of the ANN. This methodology was used for the study of the spreading of the viral epidemics propagation, in various epidemiological situations.