Hinge Regression Trees and HRT-Boost: Newton-Optimized Oblique Learning for Compact Tabular Models
Researchers have developed Hinge Regression Trees (HRT) and HRT-Boost, a new framework for creating compact tabular models. This approach reframes oblique split optimization as a nonlinear least-squares problem, allowing for efficient node-level optimization. The HRT-Boost ensemble method further enhances performance with stage-wise functional gradient descent, showing competitive results against existing baselines and producing smaller models. AI
IMPACT Introduces a novel method for building more compact and efficient tabular models, potentially improving performance in data analysis tasks.