Dynamic scaling for low-precision learning.

Ruobing Han, Min Si, James Demmel, Yang You

Browse the full PPoPP paper archive.