Recurrent neural network training with preconditioned stochastic gradient descent | Read Paper on Bytez