Artificial neural networks in time domain electromagnetics

The application of artificial neural networks (ANN) is considered to the problems of electromagnetics in time domain. Amplitudes of electromagnetic field in definite points of observation in different moments of time are proposed for using as initial values for an input layer of neural network. As examples the different structures of ANN were applied to solve the problem of determination of layered medium parameters by reflected transient field analysis. The incident field is plane wave with Gaussian time dependence. The stability of the determination of medium geometrical and electrical characteristics is studied for the case of presence of experimental errors, outer interferences, and instabilities of medium parameters.