Reviews: DECOrrelated feature space partitioning for distributed sparse regression

The paper presents a feature-wise partitioning approach for distributed sparse regression. Unfortunately, the results are rather incremental for the level of NIPS, as the result only holds for random design matrices, and the paper in its current form lacks discussion of several lines of related work and experimental baselines. While I do definitely like the conceptual idea of the partitioning followed by de-correlation, the presented theory falls short of expectations as it only holds for random design matrices. The paper however does not clearly explain the novelty and differences over [9]. Also, in addition to [9], relations to related work [B,C] are not sufficiently discussed in the current version.