Learning Morphisms with Gauss-Newton Approximation for Growing Networks