df_ppr

<p>Computes the gradient of the loss of a smoother fit to projected covariates and corresponding response values.
  Only used within optimisation of projection. Not intended for alternative use.</p>

internal

Implements the method of Hofmeyr, D.P. (2021) <DOI:10.1109/TPAMI.2019.2930501> for fast evaluation of univariate kernel smoothers based on recursive computations.
Applications to the basic problems of density and regression function estimation are provided, as well as some projection pursuit methods
for which the objective is based on non-parametric functionals of the projected density, or conditional density of a response given projected
covariates.
The package is accompanied by an instructive paper in the Journal of Statistical Software <doi:10.18637/jss.v101.i03>.

David Hofmeyr

df_ppr: Gradient of the projection index for projection pursuit regression

Description

Arguments