Explicit forms of interpolating cubic splines and data smoothing

Abstract

We express the interpolating cubic splines of class $C^2$ in their new, explicit forms. We construct the desired forms, the spline's Hermitian and B-spline representations for both equidistant and arbitrary nodes. During this process we demonstrate an innovative way to compute the inverse of a special class of tridiagonal matrices. Afterward, we propose the corresponding interpolating spline based linear regression models with easily interpretable coefficients suitable for smoothing data of complex structures.