UDC 681.325.5
MULTI-LEVEL NEURAL NETWORK APPROACH TO STRUCTURES SYNTHESIS OF FUNCTION INFORMATION CONVERTERS
A. V. Antonenko, Ph.D. (technical sciences), associate professor, ASU department, RSREU; This email address is being protected from spambots. You need JavaScript enabled to view it.
S. V. Chelebaev, Ph.D. (technical sciences), associate professor, ASU department, RSREU; This email address is being protected from spambots. You need JavaScript enabled to view it.
Y. A. Chelebaeva, undergraduate, engineer, MNEL department, RSREU; This email address is being protected from spambots. You need JavaScript enabled to view it.
Structures synthesis of function converters of information form as a component of information measuring control system is considered. The purpose of operation is reviewing the process and synthesis stages of structures of function converters using multi-level neural network description from a problem definition to finite implementation in hardware description language VHDL. A functional converter is presented in the form of integral neural network device. At the stage of problem definition basic data in the form of characteristic of conversion, criteria of end device structure optimization are set. During the synthesis the device is considered at three levels of neural network description: the level of a mathematical model of neuronet, the level of neurons converters, the level of standard neural network operations. The example of synthesis of non-linear converter of time slot in a digital code is given where the sequence of actions according to the level of a mathematical model of neuronet and the level of neurons converters is shown. Dependences of relative error of conversion on the number of weight coefficients of a network are constructed.
Keyword: synthesis, system, functional converter, neural network, neuron, activation function, time slot, digital code.