NeurIPS 2019
Sun Dec 8th through Sat the 14th, 2019 at Vancouver Convention Center
Paper ID:5380
Title:Elliptical Perturbations for Differential Privacy


		
This is a purely theoretical paper that includes results about when it is and is not possible to achieve differential privacy in infinite-dimensional spaces. The reviewer scores were 6, 7, 8; one reviewer increased their score during the discussion phase. The reviewers found the theoretical contributions insightful, significant, and original. Some critiques were raised, including accessibility to an audience not well versed in measure theory, and lack of concrete examples or applications. The author response answered a number of specific questions and promised to provide more exposition and concrete examples to address some of the reviewer concerns, which the AC encourages them to do. Overall, the consensus is that this is a good paper.