This email address is being protected from spambots. You need JavaScript enabled to view it.
 
+7 (4912) 72-03-73
 
Интернет-портал РГРТУ: http://rsreu.ru

UDC 004.383.8.032.26

DETERMINATION OF ARTIFICIAL NEURAL NETWORKS TOLERANCES BASED ON NANOMEMRISTORS

S. N. Danilin, PhD (technical sciences), assistant professor, MI VSU, Murom; This email address is being protected from spambots. You need JavaScript enabled to view it.
S. A. Shchanikov, PhD (technical sciences), assistant professor, MI VSU, Murom; This email address is being protected from spambots. You need JavaScript enabled to view it.
A. E. Sakulin, student, MI VSU, Murom, This email address is being protected from spambots. You need JavaScript enabled to view it.

A general approach to modeling and research of artificial neuron networks based on nano memristors (ANNM) as a system based on the methodology of system analysis and simulation modeling is proposed. When developing ANNM, its functional-structural decomposition was performed with the introduction of several levels of hierarchy: the system; Subsystems; Functional links; Circuit elements. A general approach is proposed to the development of methods for determining and providing quality indicators for ANNM functioning as physics-information objects. An algorithm for determining the tolerances for information parameters of the functional links of ANNM in the process of solving the synthesis problem is developed, which allows to assign tolerances to physical parameters of the means for their implementation. ANNM for detecting infocommunication signal against noise background with parameters in a given range is synthesized and investigated. The tolerances for information parameters of neurons of ANNM are determined to provide a given error in output signal for various noise parameters in input signal.

Key words: Artificial neural networks, neurocomputers, nanomemistors, operation accuracy, tolerances, signal recognition, squitter.

 Скачать статью