Metropolis-Hastings Generative Adversarial Networks

By Ryan D. Turner, Jane Hung, Yunus Saatci, Jason Yosinski. 2018

109 citations recorded by the source corpus.

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.

Subfields: Computer Vision and Pattern Recognition, Statistical and Nonlinear Physics, Generative Models, Probabilistic & Bayesian Methods

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Computer Vision and Pattern Recognition

In Statistical and Nonlinear Physics

In Generative Models

In Probabilistic & Bayesian Methods

Citations in the corpus

Your GAN is Secretly an Energy-based Model and You Should use Discriminator Driven Latent Sampling (2020, methodology): “For example, more than 600 MCMC iterations are need to obtain the most performance gain in MH-GAN [11] on real data.”
Bias Correction of Learned Generative Models using Likelihood-Free Importance Weighting (2019, neutral): “Recent concurrent works [46–48] use MCMC and rejection sampling to explicitly transform or reject the generated samples.”
Freeze the Discriminator: a Simple Baseline for Fine-Tuning GANs (2020, methodology): “…domain [24, 47, 42], using an auxiliary task to facilitate training [7, 26, 48, 49], improving an inference procedure of suboptimal models [2, 39, 29, 38], using an expressive prior distribution [13], actively choosing samples to give supervision for conditional generation [29], or actively…”
Likelihood-free MCMC with Amortized Approximate Ratio Estimators (2019, background): “…2015; Mohamed & Lakshmi-narayanan, 2016; Gutmann et al., 2017; Dutta et al., 2016; Tran et al., 2017; Brehmer et al., 2020) and is especially prominent in the area of Generative Adversarial Networks ( GAN s) (Goodfellow et al., 2014; Uehara et al., 2016; Turner et al., 2018; Azadi et al., 2018).”
Discriminator optimal transport (2019, methodology): “Recent related works in the same spirit are discriminator rejection sampling (DRS) [19] and Metropolis-a Hastings GAN (MH-GAN) [20].”
Prescribed Generative Adversarial Networks (2019, background): “Azadi et al. (2018) and Turner et al. (2018) use sampling mechanisms to correct errors of the generative distribution.”
Extracting Training Data from Diffusion Models (2023, methodology): “For the StyleGAN and MHGAN architectures, we followed the default hyper-parameters provided in the StudioGAN repository.”
Exponential Tilting of Generative Models: Improving Sample Quality by Training and Sampling from Latent Energy (2020, methodology): “Discriminator rejection sampling [2] and Metropolis-Hastings GANs [24] use the discriminator as a criterion of accepting or rejecting samples from the generator.”
Discriminator Contrastive Divergence: Semi-Amortized Generative Modeling by Exploring Energy of the Discriminator (2020, methodology): “Two recent advancements are discriminator rejection sampling(DRS) [3] and MH-GANs [36].”
Instance Selection for GANs (2020, methodology): “…of current techniques attempt to eliminate low quality samples after the model is trained, either by changing the model distribution by truncating the latent space [2, 10] or by performing some form of rejection sampling using a trained discriminator to inform the rejection process [1, 4, 26].”
Learning Latent Space Energy-Based Prior Model (2020, background): “Most of the existing papers [39, 54, 2, 13, 57, 31] assume that the latent vector follows a given simple prior distribution, such as isotropic Gaussian white noise distribution or uniform distribution.”
Bridging the Gap Between $f$-GANs and Wasserstein GANs (2019, methodology): “The learned discriminators in GANs can further be used to perform reweighting over the generated samples (Tao et al., 2018); these include rejection sampling (Azadi et al., 2018), importance sampling (Grover et al., 2019; Tao et al., 2018), and Markov chain monte carlo (Turner et al., 2018).”
Refining Deep Generative Models via Discriminator Gradient Flow (2020, methodology): “Thus far, prior work (Azadi et al., 2019; Turner et al., 2019; Tanaka, 2019; Che et al., 2020) has focused on improving samples for GANs with scalar-valued discriminators, which comprises the canonical GAN as well as recent variants, e.g., WGAN (Gulrajani et al., 2017), and SNGAN (Miyato et al.,…”
Deep Generative Modelling: A Comparative Review of VAEs, GANs, Normalizing Flows, Energy-Based and Autoregressive Models (2021, background): “which converges to the density learned by the discriminator, correcting errors learned by the generator [159], [223].”
Generalized Energy Based Models (2020, methodology): “Cifar10 LSUN CelebA ImageNet KALE-GAN 32.03 21.67 6.91 19.37 IHM 30.47 20.63 6.39 18.15 DOT 26.35 20.41 5.93 16.21 GEBM (ULA) 23.02 16.23 5.21 14.00 GEBM (KLA) 24.29 15.25 5.38 13.94 Table 2: FID scores for different sampling methods using the same trained SNGAN (ConvNet): KALE-GAN as a baseline w/o”
NCP-VAE: Variational Autoencoders with Noise Contrastive Priors (2020, background): “In spirit similar are recent works (Azadi et al., 2018; Turner et al., 2019; Che et al., 2020) that link GANs, deﬁned in the pixels space, to EBMs.”
KALE: When Energy-Based Learning Meets Adversarial Training (2020, methodology): “While there has been a recent interest in using the discriminator/energy to improve the quality of the generator/base (Azadi et al., 2019; Grover et al., 2019; Turner et al., 2019; Tanaka, 2019) (Section 5), our approach emerges naturally from the energy model we consider.”