Probabilistic-deterministic neutron-like structures: a base forperspective neurocomputers

摘要

A purely formalized approach is suggested which enables a designerto make neuron-like networks by calculating and without resorting toheuristics and indeterminancy. The resulting neuron-like networkspossess high reliability and survivability. Even when a large proportionof constituent elements (up to a fixed threshold value) stopsfunctioning, the network as a whole continues to work as if all theconstituent elements were perfect. The main organizing principles ofprobabilistic-deterministic neuron-like networks (PDNN) are presented.Principles for the construction of mathematical models of PDNNfunctioning are also examined