Algorithm for Determining Optimum Operation Tolerances of Memristor-Based Artificial Neural Networks

The article contains the results obtained by applying a general approach, which has been developed by the authors, to determining optimum values (reaching the specified level or receiving the maximum possible value) of operation tolerances of high performance memristor-based artificial neural networks (ANNM) through the use of the methodologies of system analysis and simulation modeling. Using this approach the authors have devised and programmatically implemented an algorithm aimed to solve the stated-above problem by means of varying the digit capacity of input information. The operation of a two-layer feedforward ANNM, taught to detect the squitter of a ADS-B signal with noises and interference in the background, has been used as an example during the research. In the course of the research, the authors have developed a general mathematical model of the system and a general modelling algorithm. They have also carried out simulation modelling, defined and optimized the tolerances for the parameters of neurons of the synthesized ANNM.