In a paper titled Generative adversarial networks (link in comment - sorry I don't have enough reputation to include more than 2 links), an explanation of generative adversarial net is provided as following:
I have 2 questions about the explanation of figure 1:
"G contracts in regions of high density and expands in regions of low density of p_g". I'm interpreting this to mean that the maximum of the green curve indicates region of contractions, while the region of curve flattening indicates region of increasing expansion. Why are the concepts of contractions and expansions visually represented in such a manner ?
The blue dashed line is supposed to indicate the discriminative distribution's discrimination between samples from the data generating distribution and the generative distribution. It's unclear to me how the blue dashed line represents the discrimination (why does a straight horizontal line represents the discriminator's inability to differentiate between samples taken from real data and taken from the generative distribution?)
Thank you!
