Stratified models of self-similar modular neural networks

Abstract

A stratified model for self-similar modular neural networks that includes morphological, structural, topological and parametric levels is proposed. It is shown that the morphogenesis of the morphological model is determined in the population of terminal projections of the neural network modules. Structurally regular neural networks are considered. Simplification of modular neural network models without loss of functionality is achieved using translational connections. It is shown that fast transformation algorithms (including FFT) can be described by a topological model of a structurally regular self-similar network. A linguistic model for describing the topologies of regular self-similar networks is presented. An algorithm for constructing topological models of fast algorithms is proposed. The sufficiency of the topological model for describing the complete set of fast algorithms is shown. Parametric model of modular neural network is presented. Examples are given.