Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Fully Test-time Adaptation for Tabular Data

Zhi Zhou, Kun-Yang Yu, Lan-Zhe Guo, Yu-Feng Li

TL;DR

AdaTab addresses fully test-time adaptation for tabular data under distribution shifts. The proposed FtaT framework comprises three modules—Confident Distribution Optimizer, Local Consistent Weighter, and Dynamic Model Ensembler—to handle label shift, covariate shift, and adaptation sensitivity, respectively. Empirical results on six TableShift benchmarks across three backbones show that FtaT consistently outperforms non-adaptive baselines and existing FTTA methods. The work highlights the practical importance of tabular-specific FTTA and sets the stage for deeper theoretical development and broader application to real-world tabular domains.

Abstract

Tabular data plays a vital role in various real-world scenarios and finds extensive applications. Although recent deep tabular models have shown remarkable success, they still struggle to handle data distribution shifts, leading to performance degradation when testing distributions change. To remedy this, a robust tabular model must adapt to generalize to unknown distributions during testing. In this paper, we investigate the problem of fully test-time adaptation (FTTA) for tabular data, where the model is adapted using only the testing data. We identify three key challenges: the existence of label and covariate distribution shifts, the lack of effective data augmentation, and the sensitivity of adaptation, which render existing FTTA methods ineffective for tabular data. To this end, we propose the Fully Test-time Adaptation for Tabular data, namely FTAT, which enables FTTA methods to robustly optimize the label distribution of predictions, adapt to shifted covariate distributions, and suit a variety of tasks and models effectively. We conduct comprehensive experiments on six benchmark datasets, which are evaluated using three metrics. The experimental results demonstrate that FTAT outperforms state-of-the-art methods by a margin.

Fully Test-time Adaptation for Tabular Data

TL;DR

AdaTab addresses fully test-time adaptation for tabular data under distribution shifts. The proposed FtaT framework comprises three modules—Confident Distribution Optimizer, Local Consistent Weighter, and Dynamic Model Ensembler—to handle label shift, covariate shift, and adaptation sensitivity, respectively. Empirical results on six TableShift benchmarks across three backbones show that FtaT consistently outperforms non-adaptive baselines and existing FTTA methods. The work highlights the practical importance of tabular-specific FTTA and sets the stage for deeper theoretical development and broader application to real-world tabular domains.

Abstract

Tabular data plays a vital role in various real-world scenarios and finds extensive applications. Although recent deep tabular models have shown remarkable success, they still struggle to handle data distribution shifts, leading to performance degradation when testing distributions change. To remedy this, a robust tabular model must adapt to generalize to unknown distributions during testing. In this paper, we investigate the problem of fully test-time adaptation (FTTA) for tabular data, where the model is adapted using only the testing data. We identify three key challenges: the existence of label and covariate distribution shifts, the lack of effective data augmentation, and the sensitivity of adaptation, which render existing FTTA methods ineffective for tabular data. To this end, we propose the Fully Test-time Adaptation for Tabular data, namely FTAT, which enables FTTA methods to robustly optimize the label distribution of predictions, adapt to shifted covariate distributions, and suit a variety of tasks and models effectively. We conduct comprehensive experiments on six benchmark datasets, which are evaluated using three metrics. The experimental results demonstrate that FTAT outperforms state-of-the-art methods by a margin.

Paper Structure

This paper contains 43 sections, 8 equations, 6 figures, 8 tables.

Table of Contents

  1. Introduction
  2. Problem and Analysis
  3. Problem Formulation
  4. Problem Analysis
  5. Observation 1: Covariate distribution and label distribution shifts in tabular data hinder performance of FTTA methods.
  6. Observation 2: Typical Augmentation used in test-time adaptation is ineffective for tabular data.
  7. Observation 3: Adaptation is sensitive to both tasks and models for tabular data.
  8. Observation 4: Existing FTTA methods degradates when dealing with tabular data.
  9. Methodology
  10. Confident Distribution Optimizer
  11. Local Consistent Weighter
  12. Dynamic Model Ensembler
  13. Experiments
  14. Experimental Setup
  15. Evaluation Protocol.
  16. ...and 28 more sections

Figures (6)

  • Figure 1: The label and covariate distribution shifts between training and testing in tabular data degrade the model performance. The shift degree is taken logarithm for aesthetic purposes.
  • Figure 2: The label distribution of data for which the entropy of model predictions is lower than various thresholds. The ground truth is marked with dashed line.
  • Figure 3: The performance of FtaT with different learning rates. The optimal value differs across backbones and tasks. The highest point of each line is marked by a red star.
  • Figure 4: The overall illustation of FtaT approach.
  • Figure 5: The performance of LAME, ODS, and FtaT in estimating label distribution evaluated using KL divergence.
  • ...and 1 more figures