I had thought that it only worked for 2D distributions. I am trying to implement a sliced Wasserstein autoencoder and I was wondering if my latent space can be larger than 2D.

In the WGAN-paper, an equivalent formulation of the Wasserstein distance is used. From my understanding, the Wasserstein distance is symmetric, but the version in paragraph 3 doesn't seem like it is. Is there maybe an absolute value function needed to avoid a negative sign?

I turn a directed labeled graph to histogram. for example if there is one node with label 'a', one with 'b' and two edges between them with label 'x', I turned it to axb and it's value is 2: H('axb')=2. I have some graphs that they belong to same operation but their nodes have different labels (for example if one node belongs to a process in system with pid 100, this process in another graph has different pid) I want …

I have modified a keras cyclegan keras cyclegan version of horses and zebras to the classical fer2013 face recognition file. I got some results this cyclegan trying to get some additional DISGUST (disgust is the case with fewer samples in the file) faces from NORMALS and I'm trying to modify the discriminator loss with wasserstein as stated here: But after some epochs I see that after every epoch result face is the same no matter the input face is fed. …