Random orthogonal additive filters: a solution to the vanishing/exploding gradient of deep neural networks

Open in new window