Predicting Macro Basis Functions for Method of Moments Scattering Problems Using Deep Neural Networks

In this letter, we present research on the application of deep neural networks to predicting macro basis functions for complicated computational electromagnetics problems. We provide error statistics and representative examples for networks trained on simple and complicated datasets of method of moments scattering problems. Notably, we demonstrate that the networks learn generalizable knowledge applicable to problem types on which they were not trained. We conclude that the networks produce encouraging results, especially for cross validation, and larger training datasets will improve reliability for general scattering problems.