Neuro-Impedance Function for Electromagnetic Modeling

Artificial neural networks (ANNs) can predict electromagnetic (EM) responses under different parameters quickly and accurately while avoiding time-consuming simulations. Knowledge-based neural network (KBNN) techniques incorporating previous knowledge improves model learning and speeds up development. When accessing such knowledge is difficult, the KBNN can employ neurofunction approaches. The existing neurofunction methods tend to deal with the issues of order changing and parameter discontinuity simply from a mathematical perspective. This article is the latest addition to the review of neurofunction methods, which can be summarized as neuro-impedance function (neuro-IF) EM modeling technology. Unlike previous methods, the new neuro-IF solves the problem essentially by changing the physical meaning of the function, which fundamentally avoids the issues of order changing and parameter discontinuity. This new idea further expands the modeling capability boundaries of neurofunction methods and provide a new inspiration for future research.