QuAIL: Quality-Aware Inertial Learning for Robust Training under Data Corruption
Mattia Sabella, Alberto Archetti, Pietro Pinoli, Matteo Matteucci, Cinzia Cappiello
TL;DR
Problem: tabular models suffer from non-uniform feature corruption, especially when only column-level reliability metadata is available. Approach: QuAIL introduces a quality-aware gating layer and a moving proximal anchor to regulate updates of unreliable features while letting trustworthy ones adapt, integrated into the training objective. Findings: across 50 UCI datasets under CCAR and CNAR corruption, QuAIL yields robust improvements over neural baselines and curriculum methods, with larger gains in small data settings. Significance: this demonstrates that encoding feature reliability into optimization dynamics is a practical approach to resilient, data-quality-aware tabular learning without cleaning, repair, or instance-level labels.
Abstract
Tabular machine learning systems are frequently trained on data affected by non-uniform corruption, including noisy measurements, missing entries, and feature-specific biases. In practice, these defects are often documented only through column-level reliability indicators rather than instance-wise quality annotations, limiting the applicability of many robustness and cleaning techniques. We present QuAIL, a quality-informed training mechanism that incorporates feature reliability priors directly into the learning process. QuAIL augments existing models with a learnable feature-modulation layer whose updates are selectively constrained by a quality-dependent proximal regularizer, thereby inducing controlled adaptation across features of varying trustworthiness. This stabilizes optimization under structured corruption without explicit data repair or sample-level reweighting. Empirical evaluation across 50 classification and regression datasets demonstrates that QuAIL consistently improves average performance over neural baselines under both random and value-dependent corruption, with especially robust behavior in low-data and systematically biased settings. These results suggest that incorporating feature reliability information directly into optimization dynamics is a practical and effective approach for resilient tabular learning.