Great post, enjoyed reading!

Source: Deep Learning on Medium

Great post, enjoyed reading! I think the key difference is the implementation part (with already given equations) and the theoretical part of algorithms like WGAN — researchers in DL have to understand statistics on a very deep level to come up with such a sophisticated loss function (Wasserstein metric) and make it computationally feasible (Kantorovich-Rubinstein duality). So maybe the initial cartoon is a little bit true :)