Toward accelerating deep learning at scale using specialized hardware in the datacenter | IEEE Conference Publication | IEEE Xplore