Oral
Metropolis-Hastings Generative Adversarial Networks
Ryan Turner · Jane Hung · Eric Frank · Yunus Saatchi · Jason Yosinski

Thu Jun 13th 09:35 -- 09:40 AM @ Room 101

We introduce the Metropolis-Hastings generative adversarial network (MH-GAN), which combines aspects of Markov chain Monte Carlo and GANs. The MH-GAN draws samples from the distribution implicitly defined by a GAN's discriminator-generator pair, as opposed to standard GANs which draw samples from the distribution defined by only the generator. It uses the discriminator from GAN training to build a wrapper around the generator for improved sampling. With a perfect discriminator, this wrapped generator samples from the true distribution on the data exactly even when the generator is imperfect. We demonstrate the benefits of the improved generator on multiple benchmark datasets, including CIFAR-10 and CelebA, using the DCGAN, WGAN, and progressive GAN.

Author Information

Ryan Turner (Uber AI Labs)
Jane Hung (Uber)
Eric Frank (Uber AI Labs)
Yunus Saatchi (Uber AI Labs)
Jason Yosinski (Uber Labs)

Related Events (a corresponding poster, oral, or spotlight)