NeurIPS 2020

VIME: Extending the Success of Self- and Semi-supervised Learning to Tabular Domain

Meta Review

This paper proposes a new reconstruction loss for unsupervised training of representations. This loss extends auto-encoders via a pretext task that uses the marginal distribution of features. The reviewers were unanimous in their decision to accept this paper. (Note: the one remaining score of 5 was by a reviewer who wrote that after rebuttal they are raising their score to a 6).