Smoothed Distance Kernels for MMDs and Applications in Wasserstein Gradient Flows