SYSTEMS OF FRAGMENTS OF ARTIFICIAL NEURAL NETWORKS

摘要

When we study structure of the most used artificial neural network - a multilayer perceptron and functionality of an artificial neuron, there is possibility using certain analogy with relations between descriptions of differential equations of a certain quality, in this paper is developed new approach to these subjects. In deeper view of point there can be used some concepts of locale finite trees of fragments of artificial neural networks and further to research by usage of monounary algebras and their characteristics with description and modeling systems of investigated structures of time-varying artificial neurons.