I am trying to solve a least squares problem where my independent variables are a handful of values at each of a few hundred control points, which are then interpolated to a huge number of other points via a RBFInterpolator (with neighbours <= 10), and then those interpolated values are used to compute the residuals – each point has some target values and the interpolated variables are used to generate predictions, from which I get residuals. These points are just coordinates in space, they are not on a regular grid or anything like that. The control points are a subset of the points.