Розробка системи комп’ютерного моделювання двошарової імпульсної нейронної мережі

Abstract

In this diploma work was created and studied two-layer system computer simulation of pulse neural network recognition problem signals. The pulse neural network consisted of two layers of neurons by 54 each. For software implementation was chosen combination of system software environment MATLAB and special module CSIM. To train network was used three methods: linear classification, linear regression and back propagation method. The system has better speed, accuracy and increased level of detail reflected the functioning bilayer pulse neural network compared to analog.