in style GAN, why the expectation and variance of latent layer or latent variables (after generator) can stand for the style?

Question

everyone. I just studies some knowledge about GAN and style GAN. Just as the picture shows, I wonder in style GAN, why the expectation and variance of latent layer or latent