vincentherrmann / wasserstein-notebook Goto Github PK

Hi Vincent

    I read your post "Wasserstein GAN and the Kantorovich-Rubinstein Duality". It's really excellent, but there's still something I don't understand.

In the end of the section "Strong Duality", you wrote:
(1) "The way we constructed it, we can find y_hat so that additionally b0_hat y_hat = 0."
It means we need find y_hat which can satisfy b0_hat y_hat = 0? BTW, miss a transpose symbol?

(2) "Then changing e to any number greater than 0 has to result in b0_hat y_hat > 0."
Here, should be b_hat y_hat > 0?