Redo coefficient calculation for sparse PCE-solvers
The coefficients need to be reupdated for sparse training methods, in order for the moment calculation to match up. This is mainly an issue for scikit training methods (LassoLars, SGDR(?))
The coefficients need to be reupdated for sparse training methods, in order for the moment calculation to match up. This is mainly an issue for scikit training methods (LassoLars, SGDR(?))