Adaptive Data Augmentation for Supervised Learning over Missing Data.

Real-world data is dirty, which causes serious problems in (supervised) machine learning (ML). The widely used practice in such scenario is to first repair the labeled source (a.k.a. train) data us...