Since the substance of channel extrapolation is a mapping from one parameter subspace to another, we can resort to deep learning (DL), a powerful learning architecture, to approximate such a mapping ...