Deterministic design for neural network learning: an approach based on discrepancy | IEEE Journals & Magazine | IEEE Xplore