Comments on: On Active Learning Methods for Manifold Data