In a weighted least squares method we carry least squares regressions in a pre-defined number of nearest neighbors to compute the mapping equations of the neighborhood set. We then proceed to solve these equations for the current point, each of which makes a contribution, according to selected options, to the result.