Is the loss of the discriminator in GAN the same as that in pretraining?

LantaoYu / SeqGAN

Implementation of Sequence Generative Adversarial Nets with Policy Gradient

2.08k stars 711 forks source link

Is the loss of the discriminator in GAN the same as that in pretraining? #40

Closed eduOS closed 6 years ago

eduOS commented 6 years ago

In paper they are not the same. See the function (5) in the paper.

eduOS commented 6 years ago

Got it. that in p{data} is the ground truth and that in G{\theta} is the negative target, then function (5) and function (26) are just the same. Please correct me if I made a mistake.