Large-Scale Learning with Less RAM via Randomization

   Abstract