A Neural Network Approach for the Solution of an Electromagnetic Inverse Source Problem with SVD-based Pruning

—In this paper, we face an electromagnetic inverse problem by exploiting Neural Networks (NNs). In particular, we tackle an inverse source problem and propose an iterative pruning scheme based on the use of the Singular Value Decomposition (SVD) of weight matrices and bias vectors associated to each layer to reduce the NN complexity. The performance of the approach is shown numerically.