I am thinking how to use GAN or KL Divergence as a loss function to enforce specific specific distribution on the feature space:

Let $X \sim D$ where $D$ is some distribution. Assume we know a reference asymptotic distribution $Y \sim D_2$.

We would like to find a polynomial transformation of $X \to f(X)$ such that $f(X) \sim Y \sim D_2$.

For the case of $Y \sim D_2=N(0,1)$ the network potentially will find the z-score normalization or some mapping $f$, so $f(x) \sim Y$.

I would like to get advice how to formalize this problem in order to be able to solve it using Neural Network.

Someone mentioned something like: Quantile Transforms, but I want to use neural network.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.