NeurIPS 2020

Universal Domain Adaptation through Self Supervision

Meta Review

This paper presents a domain adaptation framework that can handle arbitrary domain shift. The reviewers like the practicality of the proposed approach and the experiment results are convincing. There is a concern regarding the main theoretical basis of the proposed method (R1). I think that this is not a valid reason to reject a paper, although of course the paper would have been stronger if there was a theoretical basis for it. However, R3 also noted the lack of clear intuition regarding the proposed method. I strongly encourage the authors to at least provide more intuition. The authors have included some discussions around this in the author response. I expect them to be included and elaborated in the paper. I also encourage the authors to include other details provided in the author response based on reviewers' feedback to improve the final version of the paper.