GRAFIČKA INTERPOLACIJA JEDNOSLOJNIH I VISESLOJNIH NEURONSKIH MREŽA U MATRIČNOM LABORATORIJU/GRAPHIC INTERPOLATION OF SINGLE-LAYER AND MULTI-LAYER OF NEURON NETWORKS IN THE MATRIX LABORATORY

Abstract

Human brain counts in a completely different way from conventional digital computers. Neurons are five to six rows of size slower than digital logic. There are human natural and artificial neural networks. The artificial nets are very similar to the human brain. The model of neurons, the mathematical model and the simulation models in the Matlab program will be presented. Matlab is a suite of high-level math labs that contain a set of tools that enable the user to easily and efficiently solve certain problems. Taking into account Matlba's capabilities, we think it is an ideal solution for the implementation of artificial neural networks and the ways of implementing algorithms for learning them. By simulation, we came to the conclusion that the two-layer network is a better choice than the one with one. In the paper, two types of neural networks will be presented using ADALINE and NANR (linear and nonlinear nonlinear networks). Different number of iterations in nonlinear networks will lead to improvement of network topology up to improving output from the neural network.