Performance–energy trade-offs of deep learning convolution algorithms on ARM processors | Publicación