1

Generative models transform noise to data. However, is there a way to ensure the output of my generative model follows a specific distribution, say uniform?

More concretely, applying a scalar function $f:\mathbb{R}^n\to \mathbb{R}$ to the output of my generative model $G(z)\in\mathbb{R}^n$ is there a way to ensure that $f(G(z))$ is uniform in an interval [a,b]?

EDIT: f is a fixed function that cannot be changed. Hence, I am looking for some ways to regularize the parameters of $G$ such that $f\circ G$ is uniform.

Butters
  • 21
  • 3

1 Answers1

0

You are saying that when you train your generative model you want $f(G(z))$ to be (at least part of) the loss function. Then training will take care of the rest.

J_H
  • 1,233
  • 9
  • 12