Scalable K-FAC Training for Deep Neural Networks With Distributed Preconditioning | IEEE Journals & Magazine | IEEE Xplore