Linking Generative Adversarial Learning and Binary Classification
It provides an alternative perspective on GAN training, but is incremental as it re-derives known decision theory results.
The paper links generative adversarial training to binary classification, showing that a powerful discriminator computes an f-divergence between real and generated samples, with implications for training f-GANs through discriminator loss design.
In this note, we point out a basic link between generative adversarial (GA) training and binary classification -- any powerful discriminator essentially computes an (f-)divergence between real and generated samples. The result, repeatedly re-derived in decision theory, has implications for GA Networks (GANs), providing an alternative perspective on training f-GANs by designing the discriminator loss function.