Time domain inverse scattering by means of neural networks

We carry out an analysis of the capability of neural networks to reconstruct the dielectric and geometric properties and to localize circular cylinders starting from time-domain data and employing different configuration of the measurement system. In order to achieve, at the same time, a good accuracy in the reconstruction and a reduction in the complexity of the possible electronic components to be employed in real applications (simple data mean, usually, basic electronic components), different kinds of data extracted from the scattered electromagnetic field are tested.