Spectral Path Regression: Directional Chebyshev Harmonics for Interpretable Tabular Learning

Milo Coombs

Classical approximation bases such as Chebyshev polynomials provide principled and interpretable representations, but their multivariate tensor-product constructions scale exponent…