There is a paper that indicates Earth mover, aka Wasserstein-1 is better for convergence of GANs: https://arxiv.org/pdf/1701.07875.pdf https://arxiv.org/pdf/170

There is a paper that indicates Earth mover, aka Wasserstein-1 is better for convergence of GANs: https://arxiv.org/pdf/1701.07875.pdf https://arxiv.org/pdf/1701.07875.pdf I tried it out on one of the Udacity Deep learning assignments using the Wasserstein loss functions built into tensorflow. I was unsuccessful in my limited use. The discriminator always ‘won out’ rather than the combo finding a saddle point. I eventually got my project to work without it, and did not go back to compare against just swapping EM back in.