Reviews: Effective Parallelisation for Machine Learning

The authors have presented a parallelization algorithm for aggregating weak learners based on Radon partitioning. They present theoretical analysis to motivate the algorithm along with empirical results to support the theory. The theoretical analysis is interesting, and the empirical results demonstrate training time and/or AUC improvements over multiple baseline algorithms, on multiple datasets. The authors also preemptively and convincingly address several questions/concerns in the Evaluation and Discussion sections. Specific Notes -------------- - Line 51: "the the" - "the" - Line 79: Is this the same size math font as elsewhere?