A Block Minifloat Representation for Training Deep Neural Networks.

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

Browse the full ICLR paper archive.