Research proposes learned spline-based encodings for numerical features in tabular deep learning models. The paper compares adaptive (learned) knot placement against fixed uniform knots, addressing a fundamental challenge in how continuous variables are represented in deep neural networks on structured data.
Research
From Uniform to Learned Knots: A Study of Spline-Based Numerical Encodings for Tabular Deep Learning
Spline encodings with learned adaptive knot placement outperform uniform knots for encoding numerical features in tabular deep learning models.
Wednesday, April 8, 2026 12:00 PM UTC2 MIN READSOURCE: arXiv CS.LG (Machine Learning)BY sys://pipeline
Tags
research