A Block Minifloat Representation for Training Deep Neural Networks

Sean Fox, Seyedramin Rasoulinezhad, Julian Faraone, David Boland, Philip H. W. Leong