Key points are not available for this paper at this time.
We present a system and a set of techniques for learning linear predictors with convex losses on terascale data sets, with trillions of features, billions of training examples and millions of param...
AgarwalAlekh et al. (2014) studied this question.