NeurIPS 2019
Sun Dec 8th through Sat the 14th, 2019 at Vancouver Convention Center
Paper ID:5862
Title:Statistical Model Aggregation via Parameter Matching

The reviewers recommend accepting the paper. Their general impression is that the beta-Bernoulli process is being used in a unique way, in that each BeP selects local parameters for a group of data rather than the features present in a single observation. I agree with the reviewers that the paper can be accepted, since it's well-written and discusses a variety of applications (albeit at a very high level). However, the technical review and presentation of the model is elementary at best, and the proposed method is very simple, and it's often not obvious how it's being applied in the applications discussed. However, among borderline papers this one has the advantage of being understandable, well-written and mature in its approach to Bayesian ML.