From d624efebcbfd456edcafa18b5e5ecb6df37cda22 Mon Sep 17 00:00:00 2001 From: Steve Schmerler Date: Tue, 30 Jul 2024 23:02:30 +0200 Subject: [PATCH] DOC: add ref to sklearn's GP and KRR in rbf docs --- doc/source/written/background/rbf.rst | 12 +++++++++++- 1 file changed, 11 insertions(+), 1 deletion(-) diff --git a/doc/source/written/background/rbf.rst b/doc/source/written/background/rbf.rst index ff393802..3b68e999 100644 --- a/doc/source/written/background/rbf.rst +++ b/doc/source/written/background/rbf.rst @@ -2,10 +2,19 @@ .. _rbf: - Radial Basis Function interpolation an regression ================================================= +.. note:: + While the material here is a useful reference and the code has been used in + production, we actually recommend to use Gaussian process regression + instead, e.g. :class:`sklearn.gaussian_process.GaussianProcessRegressor`, + or, if you want to replicate :mod:`~pwtools.rbf`, then use KRR + (:class:`sklearn.kernel_ridge.KernelRidge`). See + https://github.com/elcorto/gp_playground for a detailed `comparison of GPs + and KRR + `_. + Some background information on the method implemented in :mod:`~pwtools.rbf`. For code examples, see the doc string of :class:`~pwtools.rbf.core.Rbf` and ``examples/rbf``. @@ -164,3 +173,4 @@ Other implementations --------------------- * :class:`scipy.interpolate.Rbf` +* :class:`sklearn.kernel_ridge.KernelRidge`