TangledFeatures: Robust Feature Selection in Highly Correlated Spaces
This addresses the issue of degraded performance in feature selection for models when predictors are correlated, particularly in domains like molecular analysis, though it appears incremental as it builds on existing selection techniques.
The paper tackled the problem of feature selection in correlated feature spaces by introducing TangledFeatures, a framework that identifies representative features from groups of entangled predictors, and demonstrated its effectiveness on Alanine Dipeptide, showing that selected features correspond to structurally meaningful intra-atomic distances.
Feature selection is a fundamental step in model development, shaping both predictive performance and interpretability. Yet, most widely used methods focus on predictive accuracy, and their performance degrades in the presence of correlated predictors. To address this gap, we introduce TangledFeatures, a framework for feature selection in correlated feature spaces. It identifies representative features from groups of entangled predictors, reducing redundancy while retaining explanatory power. The resulting feature subset can be directly applied in downstream models, offering a more interpretable and stable basis for analysis compared to traditional selection techniques. We demonstrate the effectiveness of TangledFeatures on Alanine Dipeptide, applying it to the prediction of backbone torsional angles and show that the selected features correspond to structurally meaningful intra-atomic distances that explain variation in these angles.