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

These methods heavily rely on the unique structure in the domain datasets (such as spatial relationships in images or semantic relationships in language).