Applications of Neural Networks to the Analysis of Dynamics of Nonlinear Spatially Extended Systems

A system based on an artificial neural network, with the structure of multilayer perceptron, was used for the recognition of the motion of a chosen, spatially extended and nonlinear system (domain wall in thin magnetic film). The dynamics of the domain wall were presented to the network in the form of binary spatio-temporal diagrams, which were based on the numerical solutions of equations of wall motion. Preliminary procession of the diagrams, with discrete Fourier transform and back propagation method for the learning of the neural network, were used. Successful recognition of spatio-temporal diagrams was achieved, which enable us to define the type of domain wall motion.