NeurIPS 2019
Sun Dec 8th through Sat the 14th, 2019 at Vancouver Convention Center
Paper ID:4653
Title:A Communication Efficient Stochastic Multi-Block Alternating Direction Method of Multipliers

Originally, the paper received one positive score and two negative scores 7/5/5 with acceptable confidence scores 3/4/3. The major issues include missing fault tolerance, weak experiments, insufficient literature review, etc. During discussion, Reviewer #5 was somwhat convinced by the authors' rebuttal and intended to raised his/her score. However, he/she was still concerned about the assumption that K=T=O(1/\eppsilon). The AC deemed that this is not a big issue. So the AC recommended acceptance. Nonetheless, the fault tolerance issue raised by Reviewers #2 and #5 was unaddressed. So the AC did not strongly support the acceptance.