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

Figure 2: The proposed self-and semi-supervised learning frameworks on exemplary tabular data. Figure 3: The proposed data corruption procedure. All three datasets are given as with separate training and testing sets. The entire US prostate cancer dataset is used as the unlabeled data. For continuous variables, we report the median value with 25% and 75% percentiles.