Distribution Shift Aware Neural Tabular Learning
This addresses the challenge of distribution shifts in tabular data for machine learning practitioners, representing an incremental improvement with a novel method for a known bottleneck.
The paper tackles the problem of tabular learning performance deteriorating under distribution shifts between training and testing data by proposing the SAFT framework, which reframes tabular learning into a continuous representation-generation paradigm and integrates mechanisms for robustness, resulting in consistent outperformance of prior methods in robustness, effectiveness, and generalization under diverse real-world shifts.
Tabular learning transforms raw features into optimized spaces for downstream tasks, but its effectiveness deteriorates under distribution shifts between training and testing data. We formalize this challenge as the Distribution Shift Tabular Learning (DSTL) problem and propose a novel Shift-Aware Feature Transformation (SAFT) framework to address it. SAFT reframes tabular learning from a discrete search task into a continuous representation-generation paradigm, enabling differentiable optimization over transformed feature sets. SAFT integrates three mechanisms to ensure robustness: (i) shift-resistant representation via embedding decorrelation and sample reweighting, (ii) flatness-aware generation through suboptimal embedding averaging, and (iii) normalization-based alignment between training and test distributions. Extensive experiments show that SAFT consistently outperforms prior tabular learning methods in terms of robustness, effectiveness, and generalization ability under diverse real-world distribution shifts.