SupplementaryMaterialsfor DeepDiffusion-Invariant WassersteinDistributionalClassification 1 MathematicalBackgrounds

Neural Information Processing Systems 

Therefore, we always setρ = 1in this paper.