Efficient hardware acceleration of deep neural networks via arithmetic complexity reduction | Publicación