A Layer-Wise Natural Gradient Optimizer for Training Deep Neural Networks | Read Paper on Bytez