When, Where, and How: Adaptive Binning for Tabular Self-Supervised Learning

2026-06-18

AI Summary

This work addresses the challenge of limited label availability in medical tabular data, which hinders the application of deep learning, and the inadequacy of existing self-supervised methods that rely on fixed global binning strategies, thereby neglecting feature heterogeneity. The authors propose an adaptive binning approach that uniquely integrates curriculum learning with neural spectral bias, employing a coarse-to-fine dynamic discretization strategy at the feature level. This framework jointly optimizes value-space compactness and representation-space consistency during training while unifying categorical reconstruction with ordinal supervision for numerical features. Evaluated across multiple public medical tabular datasets, the method significantly improves both linear probing and fine-tuning performance without requiring manual binning adjustments, and establishes a standardized benchmark for self-supervised learning on medical tabular data.

Abstract

Medical tabular data are ubiquitous in clinical research, but deep learning for tables remains underexplored because reliable labels often require costly expert adjudication, even though structured clinical variables are routinely available in tabular form. Self-supervised learning can leverage these unlabeled tables, and recent binning-based pretexts offer a promising inductive bias, but existing objectives fix a single global quantile discretization and apply feature-agnostic supervision. We propose Adaptive Binning, a training-adaptive discretization pretext for tabular SSL that couples discretization to learning through a feature-wise coarse-to-fine curriculum. Motivated by the spectral bias of neural networks and the principles of curriculum learning, our method progressively refines discretization per feature upon plateau detection and selects representation-aware splits to jointly improve value-space concentration and representation-space coherence. A heterogeneity-aware objective unifies categorical reconstruction with ordinal supervision for numerical features, and experiments on public medical tabular datasets under unified evaluation protocols show consistent gains for linear probing and fine-tuning without dataset-specific discretization tuning. We further introduce a medical tabular SSL benchmark with standardized protocols to support reproducible progress in this underexplored domain. Our code is available at https://github.com/labhai/Adaptive-Binning.

Problem

• tabular data
• self-supervised learning
• adaptive binning
• medical data
• discretization

Innovation

• Adaptive Binning
• Tabular Self-Supervised Learning
• Curriculum Learning
• Representation-Aware Discretization
• Medical Tabular Data