NeurIPS 2020

A Dictionary Approach to Domain-Invariant Learning in Deep Networks

Meta Review

All four reviewers agree that the proposed method for domain adaptation is simple and effective, in particular with regards to its plug-and-play nature. Some earlier concerns were raised in regards to weak baselines, lack of ablations and some concerns regarding reproducibility (due to lack of details) which the rebuttal seems to have addressed convincingly (in particular adding comparisons to DSN, CAN, Advent), with one reviewer increasing his score accordingly. I trust these will be included in the final version of the paper and am thus happy to support acceptance of this paper.