Metropolis-Hastings Generative Adversarial Networks
arXiv:1811.11357
Abstract
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 only by 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.
References in corpus (7)
- Unsupervised Representation Learning with Deep Convolutional Generative Adversarial Networks
- Progressive Growing of GANs for Improved Quality, Stability, and Variation
- Improved Training of Wasserstein GANs
- Density estimation using Real NVP
- Adversarially Learned Inference
- A-NICE-MC: Adversarial Training for MCMC
- Metropolis-Hastings view on variational inference and adversarial training
Cited by in corpus (8)
- Prescribed Generative Adversarial Networks
- Likelihood-free MCMC with Amortized Approximate Ratio Estimators
- Fair Generative Modeling via Weak Supervision
- Discriminator Contrastive Divergence: Semi-Amortized Generative Modeling by Exploring Energy of the Discriminator
- Exponential Tilting of Generative Models: Improving Sample Quality by Training and Sampling from Latent Energy
- Variational Diffusion Autoencoders with Random Walk Sampling
- Collaborative Sampling in Generative Adversarial Networks
- The Implicit Metropolis-Hastings Algorithm