Tag
This paper proposes Adaptive Binning, a learning-coupled feature-wise coarse-to-fine curriculum for tabular self-supervised learning that adaptively discretizes features, improving representations on medical datasets and establishing a unified benchmark.