NeurIPS 2020

MESA: Boost Ensemble Imbalanced Learning with MEta-SAmpler

Meta Review

Three referees support accept and one indicates reject. Following the author rebuttal the discussion was limited to two accept reviewers who both indicated that the paper is borderline (unsure if the paper was good enough, would not mind if rejected). This seems to indicate more of a score of 6 than the 7 they actually gave. However, I favour accepting the paper because it is an important research area and few general purpose ensemble techniques exist to tackle the problem. One could therefore imagine further work building on this.