“…It can be noted that in most works (e.g., [2,4,[10][11][12][13][14]), PINNs do not take system parameters (i.e., geometries, field sources, material properties) as an input and therefore they must be retrained, eventually taking advantage of transfer learning whenever the system parameters in the model must be changed. However, a few exceptions are reported in [8], [9], [15] where, once trained, PINNs could provide the solution of a class of direct field problems.…”